Compare commits

..

No commits in common. "main" and "v0.2.31" have entirely different histories.

49 changed files with 1062 additions and 2609 deletions

View File

@ -1,2 +0,0 @@
[build]
rustflags = ["--cfg", "tokio_unstable"]

View File

@ -8,4 +8,4 @@ jobs:
add-assignee:
runs-on: ubuntu-latest
steps:
- uses: kentaro-m/auto-assign-action@9f6dbe84a80c6e7639d1b9698048b201052a2a94
- uses: kentaro-m/auto-assign-action@7ae38e468e64dec0af17820972bc4915aa511ec2

View File

@ -26,8 +26,6 @@ jobs:
- name: Install Protoc
uses: arduino/setup-protoc@v2
with:
repo-token: ${{ secrets.GH_TOKEN }}
- name: Install Rust toolchain
uses: dtolnay/rust-toolchain@stable
@ -57,8 +55,6 @@ jobs:
- name: Install Protoc
uses: arduino/setup-protoc@v2
with:
repo-token: ${{ secrets.GH_TOKEN }}
- name: Install Rust toolchain
uses: dtolnay/rust-toolchain@stable

View File

@ -86,7 +86,7 @@ jobs:
cache-to: type=local,dest=/tmp/.buildx-cache-new
- name: Run Trivy vulnerability scanner in tarball mode
uses: aquasecurity/trivy-action@dc5a429b52fcf669ce959baa2c2dd26090d2a6c4
uses: aquasecurity/trivy-action@6c175e9c4083a92bbca2f9724c8a5e33bc2d97a5
with:
image-ref: dragonflyoss/client:${{ steps.get_version.outputs.VERSION }}
severity: 'CRITICAL,HIGH'
@ -94,7 +94,7 @@ jobs:
output: 'trivy-results.sarif'
- name: Upload Trivy scan results to GitHub Security tab
uses: github/codeql-action/upload-sarif@76621b61decf072c1cee8dd1ce2d2a82d33c17ed
uses: github/codeql-action/upload-sarif@ff0a06e83cb2de871e5a09832bc6a81e7276941f
with:
sarif_file: 'trivy-results.sarif'
@ -181,7 +181,7 @@ jobs:
cache-to: type=local,dest=/tmp/.buildx-cache-new
- name: Run Trivy vulnerability scanner in tarball mode
uses: aquasecurity/trivy-action@dc5a429b52fcf669ce959baa2c2dd26090d2a6c4
uses: aquasecurity/trivy-action@6c175e9c4083a92bbca2f9724c8a5e33bc2d97a5
with:
image-ref: dragonflyoss/client:${{ steps.get_version.outputs.VERSION }}-debug
severity: 'CRITICAL,HIGH'
@ -189,7 +189,7 @@ jobs:
output: 'trivy-results.sarif'
- name: Upload Trivy scan results to GitHub Security tab
uses: github/codeql-action/upload-sarif@76621b61decf072c1cee8dd1ce2d2a82d33c17ed
uses: github/codeql-action/upload-sarif@ff0a06e83cb2de871e5a09832bc6a81e7276941f
with:
sarif_file: 'trivy-results.sarif'
@ -276,7 +276,7 @@ jobs:
cache-to: type=local,dest=/tmp/.buildx-cache-new
- name: Run Trivy vulnerability scanner in tarball mode
uses: aquasecurity/trivy-action@dc5a429b52fcf669ce959baa2c2dd26090d2a6c4
uses: aquasecurity/trivy-action@6c175e9c4083a92bbca2f9724c8a5e33bc2d97a5
with:
image-ref: dragonflyoss/dfinit:${{ steps.get_version.outputs.VERSION }}
severity: 'CRITICAL,HIGH'
@ -284,7 +284,7 @@ jobs:
output: 'trivy-results.sarif'
- name: Upload Trivy scan results to GitHub Security tab
uses: github/codeql-action/upload-sarif@76621b61decf072c1cee8dd1ce2d2a82d33c17ed
uses: github/codeql-action/upload-sarif@ff0a06e83cb2de871e5a09832bc6a81e7276941f
with:
sarif_file: 'trivy-results.sarif'

View File

@ -15,15 +15,13 @@ jobs:
- name: Checkout code
uses: actions/checkout@v4
- name: Rust cache
- name: Rust cache
uses: Swatinem/rust-cache@v2
with:
cache-on-failure: true
- name: Install Protoc
uses: arduino/setup-protoc@v2
with:
repo-token: ${{ secrets.GH_TOKEN }}
- name: Install Rust toolchain
uses: dtolnay/rust-toolchain@stable

View File

@ -13,7 +13,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- name: PR impact specified
uses: mheap/github-action-required-labels@8afbe8ae6ab7647d0c9f0cfa7c2f939650d22509 # v5.5
uses: mheap/github-action-required-labels@388fd6af37b34cdfe5a23b37060e763217e58b03 # v5.5
with:
mode: exactly
count: 1

View File

@ -52,7 +52,7 @@ jobs:
target: ${{ matrix.target }}
- name: Install cargo-deb
uses: taiki-e/cache-cargo-install-action@b33c63d3b3c85540f4eba8a4f71a5cc0ce030855
uses: taiki-e/cache-cargo-install-action@4d586f211d9b0bca9e7b59e57e2a0febf36c0929
with:
# Don't upgrade cargo-deb, refer to https://github.com/kornelski/cargo-deb/issues/169.
tool: cargo-deb@2.10.0
@ -119,7 +119,7 @@ jobs:
contents: write
steps:
- name: Download Release Artifacts
uses: actions/download-artifact@v5
uses: actions/download-artifact@v4
with:
path: releases
pattern: release-*

645
Cargo.lock generated

File diff suppressed because it is too large Load Diff

View File

@ -12,7 +12,7 @@ members = [
]
[workspace.package]
version = "1.0.10"
version = "0.2.31"
authors = ["The Dragonfly Developers"]
homepage = "https://d7y.io/"
repository = "https://github.com/dragonflyoss/client.git"
@ -22,14 +22,14 @@ readme = "README.md"
edition = "2021"
[workspace.dependencies]
dragonfly-client = { path = "dragonfly-client", version = "1.0.10" }
dragonfly-client-core = { path = "dragonfly-client-core", version = "1.0.10" }
dragonfly-client-config = { path = "dragonfly-client-config", version = "1.0.10" }
dragonfly-client-storage = { path = "dragonfly-client-storage", version = "1.0.10" }
dragonfly-client-backend = { path = "dragonfly-client-backend", version = "1.0.10" }
dragonfly-client-util = { path = "dragonfly-client-util", version = "1.0.10" }
dragonfly-client-init = { path = "dragonfly-client-init", version = "1.0.10" }
dragonfly-api = "2.1.57"
dragonfly-client = { path = "dragonfly-client", version = "0.2.31" }
dragonfly-client-core = { path = "dragonfly-client-core", version = "0.2.31" }
dragonfly-client-config = { path = "dragonfly-client-config", version = "0.2.31" }
dragonfly-client-storage = { path = "dragonfly-client-storage", version = "0.2.31" }
dragonfly-client-backend = { path = "dragonfly-client-backend", version = "0.2.31" }
dragonfly-client-util = { path = "dragonfly-client-util", version = "0.2.31" }
dragonfly-client-init = { path = "dragonfly-client-init", version = "0.2.31" }
dragonfly-api = "=2.1.39"
thiserror = "2.0"
futures = "0.3.31"
reqwest = { version = "0.12.4", features = [
@ -41,12 +41,11 @@ reqwest = { version = "0.12.4", features = [
"brotli",
"zstd",
"deflate",
"blocking",
] }
reqwest-middleware = "0.4"
rcgen = { version = "0.12.1", features = ["x509-parser"] }
hyper = { version = "1.6", features = ["full"] }
hyper-util = { version = "0.1.16", features = [
hyper-util = { version = "0.1.12", features = [
"client",
"client-legacy",
"tokio",
@ -62,7 +61,7 @@ rustls = { version = "0.22.4", features = ["tls12"] }
rustls-pki-types = "1.12.0"
rustls-pemfile = "2.2.0"
sha2 = "0.10"
crc32fast = "1.5.0"
crc32fast = "1.4.2"
uuid = { version = "1.16", features = ["v4"] }
hex = "0.4"
rocksdb = "0.22.0"
@ -71,12 +70,12 @@ serde_yaml = "0.9"
http = "1"
tonic = { version = "0.12.2", features = ["tls"] }
tonic-reflection = "0.12.3"
tokio = { version = "1.47.1", features = ["full", "tracing"] }
tokio-util = { version = "0.7.16", features = ["full"] }
tokio = { version = "1.45.1", features = ["full"] }
tokio-util = { version = "0.7.15", features = ["full"] }
tokio-stream = "0.1.17"
validator = { version = "0.16", features = ["derive"] }
warp = "0.3.5"
headers = "0.4.1"
headers = "0.4.0"
regex = "1.11.1"
humantime = "2.1.0"
prost-wkt-types = "0.6"
@ -91,22 +90,20 @@ opendal = { version = "0.48.0", features = [
"services-cos",
"services-webhdfs",
] }
clap = { version = "4.5.45", features = ["derive"] }
clap = { version = "4.5.38", features = ["derive"] }
anyhow = "1.0.98"
toml_edit = "0.22.26"
toml = "0.8.23"
toml = "0.8.22"
bytesize = { version = "1.3.3", features = ["serde"] }
bytesize-serde = "0.2.1"
percent-encoding = "2.3.1"
tempfile = "3.20.0"
tokio-rustls = "0.25.0-alpha.4"
serde_json = "1.0.142"
serde_json = "1.0.140"
lru = "0.12.5"
fs2 = "0.4.3"
lazy_static = "1.5"
bytes = "1.10"
local-ip-address = "0.6.5"
sysinfo = { version = "0.32.1", default-features = false, features = ["component", "disk", "network", "system", "user"] }
[profile.release]
opt-level = 3

View File

@ -30,3 +30,7 @@ Join the conversation and help the community.
You should check out our
[CONTRIBUTING](./CONTRIBUTING.md) and develop the project together.
## License
[![FOSSA Status](https://app.fossa.com/api/projects/git%2Bgithub.com%2Fdragonflyoss%2Fclient.svg?type=large)](https://app.fossa.com/projects/git%2Bgithub.com%2Fdragonflyoss%2Fclient?ref=badge_large)

View File

@ -1,4 +1,4 @@
FROM public.ecr.aws/docker/library/rust:1.85.0 AS builder
FROM public.ecr.aws/docker/library/rust:1.82.0 AS builder
WORKDIR /app/client
@ -7,7 +7,6 @@ RUN apt-get update && apt-get install -y \
&& rm -rf /var/lib/apt/lists/*
COPY Cargo.toml Cargo.lock ./
COPY .cargo ./cargo
COPY dragonfly-client/Cargo.toml ./dragonfly-client/Cargo.toml
COPY dragonfly-client/src ./dragonfly-client/src
@ -35,13 +34,7 @@ COPY dragonfly-client-util/src ./dragonfly-client-util/src
COPY dragonfly-client-init/Cargo.toml ./dragonfly-client-init/Cargo.toml
COPY dragonfly-client-init/src ./dragonfly-client-init/src
ARG TARGETPLATFORM
RUN case "${TARGETPLATFORM}" in \
"linux/arm64") export JEMALLOC_SYS_WITH_LG_PAGE=16;; \
esac && \
cargo build --release --verbose --bin dfget --bin dfdaemon --bin dfcache
RUN cargo install tokio-console --locked --root /usr/local
RUN cargo build --release --verbose --bin dfget --bin dfdaemon --bin dfcache
FROM public.ecr.aws/docker/library/alpine:3.20 AS health
@ -59,21 +52,17 @@ RUN if [ "$(uname -m)" = "ppc64le" ]; then \
FROM public.ecr.aws/docker/library/golang:1.23.0-alpine3.20 AS pprof
RUN go install github.com/google/pprof@latest
RUN go install github.com/fullstorydev/grpcurl/cmd/grpcurl@latest
FROM public.ecr.aws/debian/debian:bookworm-slim
RUN apt-get update && apt-get install -y --no-install-recommends iperf3 fio curl \
iotop sysstat bash-completion procps apache2-utils ca-certificates binutils \
dnsutils iputils-ping llvm graphviz lsof strace dstat net-tools \
RUN apt-get update && apt-get install -y --no-install-recommends curl bash-completion procps infiniband-diags ibverbs-utils \
apache2-utils ca-certificates binutils dnsutils iputils-ping llvm dstat sysstat net-tools \
&& rm -rf /var/lib/apt/lists/*
COPY --from=builder /app/client/target/release/dfget /usr/local/bin/dfget
COPY --from=builder /app/client/target/release/dfdaemon /usr/local/bin/dfdaemon
COPY --from=builder /app/client/target/release/dfcache /usr/local/bin/dfcache
COPY --from=builder /usr/local/bin/tokio-console /usr/local/bin/
COPY --from=pprof /go/bin/pprof /bin/pprof
COPY --from=pprof /go/bin/grpcurl /bin/grpcurl
COPY --from=health /bin/grpc_health_probe /bin/grpc_health_probe
ENTRYPOINT ["/usr/local/bin/dfdaemon"]

View File

@ -1,4 +1,4 @@
FROM public.ecr.aws/docker/library/rust:1.85.0 AS builder
FROM public.ecr.aws/docker/library/rust:1.82.0 AS builder
WORKDIR /app/client
@ -7,7 +7,6 @@ RUN apt-get update && apt-get install -y \
&& rm -rf /var/lib/apt/lists/*
COPY Cargo.toml Cargo.lock ./
COPY .cargo ./cargo
COPY dragonfly-client/Cargo.toml ./dragonfly-client/Cargo.toml
COPY dragonfly-client/src ./dragonfly-client/src
@ -35,15 +34,10 @@ COPY dragonfly-client-util/src ./dragonfly-client-util/src
COPY dragonfly-client-init/Cargo.toml ./dragonfly-client-init/Cargo.toml
COPY dragonfly-client-init/src ./dragonfly-client-init/src
ARG TARGETPLATFORM
RUN case "${TARGETPLATFORM}" in \
"linux/arm64") export JEMALLOC_SYS_WITH_LG_PAGE=16;; \
esac && \
cargo build --verbose --bin dfget --bin dfdaemon --bin dfcache
RUN cargo build --verbose --bin dfget --bin dfdaemon --bin dfcache
RUN cargo install flamegraph --root /usr/local
RUN cargo install bottom --locked --root /usr/local
RUN cargo install tokio-console --locked --root /usr/local
FROM public.ecr.aws/docker/library/alpine:3.20 AS health
@ -61,13 +55,12 @@ RUN if [ "$(uname -m)" = "ppc64le" ]; then \
FROM public.ecr.aws/docker/library/golang:1.23.0-alpine3.20 AS pprof
RUN go install github.com/google/pprof@latest
RUN go install github.com/fullstorydev/grpcurl/cmd/grpcurl@latest
FROM public.ecr.aws/debian/debian:bookworm-slim
RUN apt-get update && apt-get install -y --no-install-recommends iperf3 fio curl infiniband-diags ibverbs-utils \
RUN apt-get update && apt-get install -y --no-install-recommends iperf3 fio wget curl infiniband-diags ibverbs-utils \
iotop sysstat bash-completion procps apache2-utils ca-certificates binutils bpfcc-tools \
dnsutils iputils-ping vim linux-perf llvm lsof socat strace dstat net-tools \
dnsutils iputils-ping vim linux-perf llvm graphviz lsof socat strace dstat net-tools \
&& rm -rf /var/lib/apt/lists/*
COPY --from=builder /app/client/target/debug/dfget /usr/local/bin/dfget
@ -75,9 +68,7 @@ COPY --from=builder /app/client/target/debug/dfdaemon /usr/local/bin/dfdaemon
COPY --from=builder /app/client/target/debug/dfcache /usr/local/bin/dfcache
COPY --from=builder /usr/local/bin/flamegraph /usr/local/bin/
COPY --from=builder /usr/local/bin/btm /usr/local/bin/
COPY --from=builder /usr/local/bin/tokio-console /usr/local/bin/
COPY --from=pprof /go/bin/pprof /bin/pprof
COPY --from=pprof /go/bin/grpcurl /bin/grpcurl
COPY --from=health /bin/grpc_health_probe /bin/grpc_health_probe
ENTRYPOINT ["/usr/local/bin/dfdaemon"]

View File

@ -1,4 +1,4 @@
FROM public.ecr.aws/docker/library/rust:1.85.0 AS builder
FROM public.ecr.aws/docker/library/rust:1.82.0 AS builder
RUN apt-get update && apt-get install -y \
openssl libclang-dev pkg-config protobuf-compiler \
@ -7,7 +7,6 @@ RUN apt-get update && apt-get install -y \
WORKDIR /app/client
COPY Cargo.toml Cargo.lock ./
COPY .cargo ./cargo
COPY dragonfly-client/Cargo.toml ./dragonfly-client/Cargo.toml
COPY dragonfly-client/src ./dragonfly-client/src
@ -35,11 +34,7 @@ COPY dragonfly-client-util/src ./dragonfly-client-util/src
COPY dragonfly-client-init/Cargo.toml ./dragonfly-client-init/Cargo.toml
COPY dragonfly-client-init/src ./dragonfly-client-init/src
ARG TARGETPLATFORM
RUN case "${TARGETPLATFORM}" in \
"linux/arm64") export JEMALLOC_SYS_WITH_LG_PAGE=16;; \
esac && \
cargo build --release --verbose --bin dfinit
RUN cargo build --release --verbose --bin dfinit
FROM public.ecr.aws/debian/debian:bookworm-slim

View File

@ -27,11 +27,11 @@ percent-encoding.workspace = true
futures.workspace = true
reqwest-retry = "0.7"
reqwest-tracing = "0.5"
libloading = "0.8.8"
libloading = "0.8.7"
[dev-dependencies]
tempfile.workspace = true
wiremock = "0.6.4"
wiremock = "0.6.3"
rustls-pki-types.workspace = true
rustls-pemfile.workspace = true
hyper.workspace = true

View File

@ -50,22 +50,11 @@ impl HTTP {
.with_custom_certificate_verifier(NoVerifier::new())
.with_no_client_auth();
// Disable automatic compression to prevent double-decompression issues.
//
// Problem scenario:
// 1. Origin server supports gzip and returns "content-encoding: gzip" header.
// 2. Backend decompresses the response and stores uncompressed content to disk.
// 3. When user's client downloads via dfdaemon proxy, the original "content-encoding: gzip".
// header is forwarded to it.
// 4. User's client attempts to decompress the already-decompressed content, causing errors.
//
// Solution: Disable all compression formats (gzip, brotli, zstd, deflate) to ensure
// we receive and store uncompressed content, eliminating the double-decompression issue.
let client = reqwest::Client::builder()
.no_gzip()
.no_brotli()
.no_zstd()
.no_deflate()
.gzip(true)
.brotli(true)
.zstd(true)
.deflate(true)
.use_preconfigured_tls(client_config_builder)
.pool_max_idle_per_host(super::POOL_MAX_IDLE_PER_HOST)
.tcp_keepalive(super::KEEP_ALIVE_INTERVAL)
@ -99,22 +88,11 @@ impl HTTP {
.with_root_certificates(root_cert_store)
.with_no_client_auth();
// Disable automatic compression to prevent double-decompression issues.
//
// Problem scenario:
// 1. Origin server supports gzip and returns "content-encoding: gzip" header.
// 2. Backend decompresses the response and stores uncompressed content to disk.
// 3. When user's client downloads via dfdaemon proxy, the original "content-encoding: gzip".
// header is forwarded to it.
// 4. User's client attempts to decompress the already-decompressed content, causing errors.
//
// Solution: Disable all compression formats (gzip, brotli, zstd, deflate) to ensure
// we receive and store uncompressed content, eliminating the double-decompression issue.
let client = reqwest::Client::builder()
.no_gzip()
.no_brotli()
.no_zstd()
.no_deflate()
.gzip(true)
.brotli(true)
.zstd(true)
.deflate(true)
.use_preconfigured_tls(client_config_builder)
.build()?;
@ -160,13 +138,6 @@ impl super::Backend for HTTP {
.client(request.client_cert)?
.get(&request.url)
.headers(header)
// Add Range header to ensure Content-Length is returned in response headers.
// Some servers (especially when using Transfer-Encoding: chunked,
// refer to https://developer.mozilla.org/en-US/docs/Web/HTTP/Reference/Headers/Transfer-Encoding.) may not
// include Content-Length in HEAD requests. Using "bytes=0-" requests the
// entire file starting from byte 0, forcing the server to include file size
// information in the response headers.
.header(reqwest::header::RANGE, "bytes=0-")
.timeout(request.timeout)
.send()
.await

View File

@ -23,6 +23,7 @@ use libloading::Library;
use reqwest::header::HeaderMap;
use rustls_pki_types::CertificateDer;
use std::path::Path;
use std::str::FromStr;
use std::{collections::HashMap, pin::Pin, time::Duration};
use std::{fmt::Debug, fs};
use tokio::io::{AsyncRead, AsyncReadExt};
@ -166,7 +167,7 @@ where
}
/// The File Entry of a directory, including some relevant file metadata.
#[derive(Debug, PartialEq, Eq, Hash, Clone)]
#[derive(Debug, PartialEq, Eq)]
pub struct DirEntry {
/// url is the url of the entry.
pub url: String,
@ -240,9 +241,9 @@ impl BackendFactory {
Ok(backend_factory)
}
/// unsupported_download_directory returns whether the scheme does not support directory download.
pub fn unsupported_download_directory(scheme: &str) -> bool {
scheme == http::HTTP_SCHEME || scheme == http::HTTPS_SCHEME
/// supported_download_directory returns whether the scheme supports directory download.
pub fn supported_download_directory(scheme: &str) -> bool {
object_storage::Scheme::from_str(scheme).is_ok() || scheme == hdfs::HDFS_SCHEME
}
/// build returns the backend by the scheme of the url.
@ -330,7 +331,7 @@ impl BackendFactory {
if !backend_plugin_dir.exists() {
warn!(
"skip loading plugin backends, because the plugin directory {} does not exist",
backend_plugin_dir.display()
plugin_dir.display()
);
return Ok(());
}

View File

@ -13,7 +13,6 @@ build = "build.rs"
[dependencies]
dragonfly-client-core.workspace = true
dragonfly-client-util.workspace = true
local-ip-address.workspace = true
clap.workspace = true
regex.workspace = true
serde.workspace = true
@ -29,9 +28,8 @@ bytesize-serde.workspace = true
tonic.workspace = true
rustls-pki-types.workspace = true
rcgen.workspace = true
reqwest.workspace = true
home = "0.5.11"
local-ip-address = "0.6.5"
hostname = "^0.4"
humantime-serde = "1.1.1"
serde_regex = "1.1.0"
http-serde = "2.1.1"

View File

@ -146,12 +146,6 @@ fn default_download_piece_timeout() -> Duration {
Duration::from_secs(120)
}
/// default_collected_download_piece_timeout is the default timeout for collecting one piece from the parent in the stream.
#[inline]
fn default_collected_download_piece_timeout() -> Duration {
Duration::from_secs(10)
}
/// default_download_concurrent_piece_count is the default number of concurrent pieces to download.
#[inline]
fn default_download_concurrent_piece_count() -> u32 {
@ -164,12 +158,6 @@ fn default_download_max_schedule_count() -> u32 {
5
}
/// default_tracing_path is the default tracing path for dfdaemon.
#[inline]
fn default_tracing_path() -> Option<PathBuf> {
Some(PathBuf::from("/v1/traces"))
}
/// default_scheduler_announce_interval is the default interval to announce peer to the scheduler.
#[inline]
fn default_scheduler_announce_interval() -> Duration {
@ -226,6 +214,18 @@ fn default_storage_cache_capacity() -> ByteSize {
ByteSize::mib(64)
}
/// default_seed_peer_cluster_id is the default cluster id of seed peer.
#[inline]
fn default_seed_peer_cluster_id() -> u64 {
1
}
/// default_seed_peer_keepalive_interval is the default interval to keepalive with manager.
#[inline]
fn default_seed_peer_keepalive_interval() -> Duration {
Duration::from_secs(15)
}
/// default_gc_interval is the default interval to do gc.
#[inline]
fn default_gc_interval() -> Duration {
@ -238,12 +238,6 @@ fn default_gc_policy_task_ttl() -> Duration {
Duration::from_secs(21_600)
}
/// default_gc_policy_dist_threshold is the default threshold of the disk usage to do gc.
#[inline]
fn default_gc_policy_dist_threshold() -> ByteSize {
ByteSize::default()
}
/// default_gc_policy_dist_high_threshold_percent is the default high threshold percent of the disk usage.
#[inline]
fn default_gc_policy_dist_high_threshold_percent() -> u8 {
@ -401,12 +395,6 @@ pub struct Host {
/// ip is the advertise ip of the host.
pub ip: Option<IpAddr>,
/// scheduler_cluster_id is the ID of the cluster to which the scheduler belongs.
/// NOTE: This field is used to identify the cluster to which the scheduler belongs.
/// If this flag is set, the idc, location, hostname and ip will be ignored when listing schedulers.
#[serde(rename = "schedulerClusterID")]
pub scheduler_cluster_id: Option<u64>,
}
/// Host implements Default.
@ -417,7 +405,6 @@ impl Default for Host {
location: None,
hostname: default_host_hostname(),
ip: None,
scheduler_cluster_id: None,
}
}
}
@ -487,14 +474,6 @@ pub struct Download {
#[serde(default = "default_download_piece_timeout", with = "humantime_serde")]
pub piece_timeout: Duration,
/// collected_piece_timeout is the timeout for collecting one piece from the parent in the
/// stream.
#[serde(
default = "default_collected_download_piece_timeout",
with = "humantime_serde"
)]
pub collected_piece_timeout: Duration,
/// concurrent_piece_count is the number of concurrent pieces to download.
#[serde(default = "default_download_concurrent_piece_count")]
#[validate(range(min = 1))]
@ -509,7 +488,6 @@ impl Default for Download {
parent_selector: ParentSelector::default(),
rate_limit: default_download_rate_limit(),
piece_timeout: default_download_piece_timeout(),
collected_piece_timeout: default_collected_download_piece_timeout(),
concurrent_piece_count: default_download_concurrent_piece_count(),
}
}
@ -912,6 +890,18 @@ pub struct SeedPeer {
/// kind is the type of seed peer.
#[serde(default, rename = "type")]
pub kind: HostType,
/// cluster_id is the cluster id of the seed peer cluster.
#[serde(default = "default_seed_peer_cluster_id", rename = "clusterID")]
#[validate(range(min = 1))]
pub cluster_id: u64,
/// keepalive_interval is the interval to keep alive with manager.
#[serde(
default = "default_seed_peer_keepalive_interval",
with = "humantime_serde"
)]
pub keepalive_interval: Duration,
}
/// SeedPeer implements Default.
@ -920,6 +910,8 @@ impl Default for SeedPeer {
SeedPeer {
enable: false,
kind: HostType::Normal,
cluster_id: default_seed_peer_cluster_id(),
keepalive_interval: default_seed_peer_keepalive_interval(),
}
}
}
@ -1001,33 +993,31 @@ pub struct Storage {
/// cache_capacity is the cache capacity for downloading, default is 100.
///
/// Cache storage:
/// 1. Users can preheat task by caching to memory (via CacheTask) or to disk (via Task).
/// For more details, refer to https://github.com/dragonflyoss/api/blob/main/proto/dfdaemon.proto#L174.
/// 1. Users can create preheating jobs and preheat tasks to memory and disk by setting `load_to_cache` to `true`.
/// For more details, refer to https://github.com/dragonflyoss/api/blob/main/proto/common.proto#L443.
/// 2. If the download hits the memory cache, it will be faster than reading from the disk, because there is no
/// page cache for the first read.
///
///```text
/// +--------+
/// │ Source │
/// +--------+
/// ^ ^ Preheat
/// │ │ |
/// +-----------------+ │ │ +----------------------------+
/// │ Other Peers │ │ │ │ Peer | │
/// │ │ │ │ │ v │
/// │ +----------+ │ │ │ │ +----------+ │
/// │ │ Cache |<--|----------|<-Miss--| Cache |--Hit-->|<----Download CacheTask
/// │ +----------+ │ │ │ +----------+ │
/// │ │ │ │ │
/// │ +----------+ │ │ │ +----------+ │
/// │ │ Disk |<--|----------|<-Miss--| Disk |--Hit-->|<----Download Task
/// │ +----------+ │ │ +----------+ │
/// │ │ │ ^ │
/// │ │ │ | │
/// +-----------------+ +----------------------------+
/// |
/// Preheat
///```
/// ```text
/// 1.Preheat
/// |
/// |
/// +--------------------------------------------------+
/// | | Peer |
/// | | +-----------+ |
/// | | -- Partial -->| Cache | |
/// | | | +-----------+ |
/// | v | | | |
/// | Download | Miss | |
/// | Task -->| | --- Hit ------>|<-- 2.Download
/// | | | ^ |
/// | | v | |
/// | | +-----------+ | |
/// | -- Full -->| Disk |---------- |
/// | +-----------+ |
/// | |
/// +--------------------------------------------------+
/// ```
#[serde(with = "bytesize_serde", default = "default_storage_cache_capacity")]
pub cache_capacity: ByteSize,
}
@ -1059,19 +1049,6 @@ pub struct Policy {
)]
pub task_ttl: Duration,
/// dist_threshold optionally defines a specific disk capacity to be used as the base for
/// calculating GC trigger points with `dist_high_threshold_percent` and `dist_low_threshold_percent`.
///
/// - If a value is provided (e.g., "500GB"), the percentage-based thresholds (`dist_high_threshold_percent`,
/// `dist_low_threshold_percent`) are applied relative to this specified capacity.
/// - If not provided or set to 0 (the default behavior), these percentage-based thresholds are applied
/// relative to the total actual disk space.
///
/// This allows dfdaemon to effectively manage a logical portion of the disk for its cache,
/// rather than always considering the entire disk volume.
#[serde(with = "bytesize_serde", default = "default_gc_policy_dist_threshold")]
pub dist_threshold: ByteSize,
/// dist_high_threshold_percent is the high threshold percent of the disk usage.
/// If the disk usage is greater than the threshold, dfdaemon will do gc.
#[serde(default = "default_gc_policy_dist_high_threshold_percent")]
@ -1089,7 +1066,6 @@ pub struct Policy {
impl Default for Policy {
fn default() -> Self {
Policy {
dist_threshold: default_gc_policy_dist_threshold(),
task_ttl: default_gc_policy_task_ttl(),
dist_high_threshold_percent: default_gc_policy_dist_high_threshold_percent(),
dist_low_threshold_percent: default_gc_policy_dist_low_threshold_percent(),
@ -1449,37 +1425,11 @@ pub struct Stats {
}
/// Tracing is the tracing configuration for dfdaemon.
#[derive(Debug, Clone, Validate, Deserialize)]
#[derive(Debug, Clone, Default, Validate, Deserialize)]
#[serde(default, rename_all = "camelCase")]
pub struct Tracing {
/// Protocol specifies the communication protocol for the tracing server.
/// Supported values: "http", "https", "grpc" (default: None).
/// This determines how tracing logs are transmitted to the server.
pub protocol: Option<String>,
/// endpoint is the endpoint to report tracing log, example: "localhost:4317".
pub endpoint: Option<String>,
/// path is the path to report tracing log, example: "/v1/traces" if the protocol is "http" or
/// "https".
#[serde(default = "default_tracing_path")]
pub path: Option<PathBuf>,
/// headers is the headers to report tracing log.
#[serde(with = "http_serde::header_map")]
pub headers: reqwest::header::HeaderMap,
}
/// Tracing implements Default.
impl Default for Tracing {
fn default() -> Self {
Self {
protocol: None,
endpoint: None,
path: default_tracing_path(),
headers: reqwest::header::HeaderMap::new(),
}
}
/// addr is the address to report tracing log.
pub addr: Option<String>,
}
/// Config is the configuration for dfdaemon.
@ -1989,6 +1939,11 @@ key: /etc/ssl/private/client.pem
let default_seed_peer = SeedPeer::default();
assert!(!default_seed_peer.enable);
assert_eq!(default_seed_peer.kind, HostType::Normal);
assert_eq!(default_seed_peer.cluster_id, 1);
assert_eq!(
default_seed_peer.keepalive_interval,
default_seed_peer_keepalive_interval()
);
}
#[test]
@ -1996,9 +1951,20 @@ key: /etc/ssl/private/client.pem
let valid_seed_peer = SeedPeer {
enable: true,
kind: HostType::Weak,
cluster_id: 5,
keepalive_interval: Duration::from_secs(90),
};
assert!(valid_seed_peer.validate().is_ok());
let invalid_seed_peer = SeedPeer {
enable: true,
kind: HostType::Weak,
cluster_id: 0,
keepalive_interval: Duration::from_secs(90),
};
assert!(invalid_seed_peer.validate().is_err());
}
#[test]
@ -2015,6 +1981,8 @@ key: /etc/ssl/private/client.pem
assert!(seed_peer.enable);
assert_eq!(seed_peer.kind, HostType::Super);
assert_eq!(seed_peer.cluster_id, 2);
assert_eq!(seed_peer.keepalive_interval, Duration::from_secs(60));
}
#[test]
@ -2065,18 +2033,18 @@ key: /etc/ssl/private/client.pem
fn validate_policy() {
let valid_policy = Policy {
task_ttl: Duration::from_secs(12 * 3600),
dist_threshold: ByteSize::mb(100),
dist_high_threshold_percent: 90,
dist_low_threshold_percent: 70,
};
assert!(valid_policy.validate().is_ok());
let invalid_policy = Policy {
task_ttl: Duration::from_secs(12 * 3600),
dist_threshold: ByteSize::mb(100),
dist_high_threshold_percent: 100,
dist_low_threshold_percent: 70,
};
assert!(invalid_policy.validate().is_err());
}
@ -2176,19 +2144,12 @@ key: /etc/ssl/private/client.pem
fn deserialize_tracing_correctly() {
let json_data = r#"
{
"protocol": "http",
"endpoint": "tracing.example.com",
"path": "/v1/traces",
"headers": {
"X-Custom-Header": "value"
}
"addr": "http://tracing.example.com"
}"#;
let tracing: Tracing = serde_json::from_str(json_data).unwrap();
assert_eq!(tracing.protocol, Some("http".to_string()));
assert_eq!(tracing.endpoint, Some("tracing.example.com".to_string()));
assert_eq!(tracing.path, Some(PathBuf::from("/v1/traces")));
assert!(tracing.headers.contains_key("X-Custom-Header"));
assert_eq!(tracing.addr, Some("http://tracing.example.com".to_string()));
}
#[test]

View File

@ -104,7 +104,7 @@ pub fn default_lock_dir() -> PathBuf {
/// default_plugin_dir is the default plugin directory for client.
pub fn default_plugin_dir() -> PathBuf {
#[cfg(target_os = "linux")]
return PathBuf::from("/usr/local/lib/dragonfly/plugins/");
return PathBuf::from("/var/lib/dragonfly/plugins/");
#[cfg(target_os = "macos")]
return home::home_dir().unwrap().join(".dragonfly").join("plugins");

View File

@ -23,6 +23,7 @@ tokio.workspace = true
anyhow.workspace = true
tracing.workspace = true
toml_edit.workspace = true
toml.workspace = true
url.workspace = true
tempfile.workspace = true
serde_json.workspace = true

View File

@ -91,9 +91,6 @@ async fn main() -> Result<(), anyhow::Error> {
args.log_max_files,
None,
None,
None,
None,
None,
false,
args.console,
);

View File

@ -22,16 +22,17 @@ tracing.workspace = true
prost-wkt-types.workspace = true
tokio.workspace = true
tokio-util.workspace = true
sha2.workspace = true
crc32fast.workspace = true
fs2.workspace = true
lru.workspace = true
bytes.workspace = true
bytesize.workspace = true
num_cpus = "1.17"
num_cpus = "1.0"
bincode = "1.3.3"
walkdir = "2.5.0"
[dev-dependencies]
tempfile.workspace = true
tempdir = "0.3"
criterion = "0.5"
[[bench]]

View File

@ -76,33 +76,31 @@ impl Task {
/// Cache is the cache for storing piece content by LRU algorithm.
///
/// Cache storage:
/// 1. Users can preheat task by caching to memory (via CacheTask) or to disk (via Task).
/// For more details, refer to https://github.com/dragonflyoss/api/blob/main/proto/dfdaemon.proto#L174.
/// 1. Users can create preheating jobs and preheat tasks to memory and disk by setting `load_to_cache` to `true`.
/// For more details, refer to https://github.com/dragonflyoss/api/blob/main/proto/common.proto#L443.
/// 2. If the download hits the memory cache, it will be faster than reading from the disk, because there is no
/// page cache for the first read.
///
///```text
/// +--------+
/// │ Source │
/// +--------+
/// ^ ^ Preheat
/// │ │ |
/// +-----------------+ │ │ +----------------------------+
/// │ Other Peers │ │ │ │ Peer | │
/// │ │ │ │ │ v │
/// │ +----------+ │ │ │ │ +----------+ │
/// │ │ Cache |<--|----------|<-Miss--| Cache |--Hit-->|<----Download CacheTask
/// │ +----------+ │ │ │ +----------+ │
/// │ │ │ │ │
/// │ +----------+ │ │ │ +----------+ │
/// │ │ Disk |<--|----------|<-Miss--| Disk |--Hit-->|<----Download Task
/// │ +----------+ │ │ +----------+ │
/// │ │ │ ^ │
/// │ │ │ | │
/// +-----------------+ +----------------------------+
/// |
/// Preheat
///```
/// ```text
/// 1.Preheat
/// |
/// |
/// +--------------------------------------------------+
/// | | Peer |
/// | | +-----------+ |
/// | | -- Partial -->| Cache | |
/// | | | +-----------+ |
/// | v | | | |
/// | Download | Miss | |
/// | Task -->| | --- Hit ------>|<-- 2.Download
/// | | | ^ |
/// | | v | |
/// | | +-----------+ | |
/// | -- Full -->| Disk |---------- |
/// | +-----------+ |
/// | |
/// +--------------------------------------------------+
/// ```
/// Task is the metadata of the task.
#[derive(Clone)]
pub struct Cache {

View File

@ -14,7 +14,6 @@
* limitations under the License.
*/
use bytesize::ByteSize;
use dragonfly_api::common::v2::Range;
use dragonfly_client_config::dfdaemon::Config;
use dragonfly_client_core::{Error, Result};
@ -28,7 +27,6 @@ use tokio::io::{
};
use tokio_util::io::InspectReader;
use tracing::{error, info, instrument, warn};
use walkdir::WalkDir;
/// DEFAULT_CONTENT_DIR is the default directory for store content.
pub const DEFAULT_CONTENT_DIR: &str = "content";
@ -87,39 +85,12 @@ impl Content {
/// available_space returns the available space of the disk.
pub fn available_space(&self) -> Result<u64> {
let dist_threshold = self.config.gc.policy.dist_threshold;
if dist_threshold != ByteSize::default() {
let usage_space = WalkDir::new(&self.dir)
.into_iter()
.filter_map(|entry| entry.ok())
.filter_map(|entry| entry.metadata().ok())
.filter(|metadata| metadata.is_file())
.fold(0, |acc, m| acc + m.len());
if usage_space >= dist_threshold.as_u64() {
warn!(
"usage space {} is greater than dist threshold {}, no need to calculate available space",
usage_space, dist_threshold
);
return Ok(0);
}
return Ok(dist_threshold.as_u64() - usage_space);
}
let stat = fs2::statvfs(&self.dir)?;
Ok(stat.available_space())
}
/// total_space returns the total space of the disk.
pub fn total_space(&self) -> Result<u64> {
// If the dist_threshold is set, return it directly.
let dist_threshold = self.config.gc.policy.dist_threshold;
if dist_threshold != ByteSize::default() {
return Ok(dist_threshold.as_u64());
}
let stat = fs2::statvfs(&self.dir)?;
Ok(stat.total_space())
}
@ -359,7 +330,6 @@ impl Content {
&self,
task_id: &str,
offset: u64,
expected_length: u64,
reader: &mut R,
) -> Result<WritePieceResponse> {
// Open the file and seek to the offset.
@ -394,13 +364,6 @@ impl Content {
error!("flush {:?} failed: {}", task_path, err);
})?;
if length != expected_length {
return Err(Error::Unknown(format!(
"expected length {} but got {}",
expected_length, length
)));
}
// Calculate the hash of the piece.
Ok(WritePieceResponse {
length,
@ -581,7 +544,6 @@ impl Content {
&self,
task_id: &str,
offset: u64,
expected_length: u64,
reader: &mut R,
) -> Result<WritePieceResponse> {
// Open the file and seek to the offset.
@ -616,13 +578,6 @@ impl Content {
error!("flush {:?} failed: {}", task_path, err);
})?;
if length != expected_length {
return Err(Error::Unknown(format!(
"expected length {} but got {}",
expected_length, length
)));
}
// Calculate the hash of the piece.
Ok(WritePieceResponse {
length,
@ -670,11 +625,11 @@ pub fn calculate_piece_range(offset: u64, length: u64, range: Option<Range>) ->
mod tests {
use super::*;
use std::io::Cursor;
use tempfile::tempdir;
use tempdir::TempDir;
#[tokio::test]
async fn test_create_task() {
let temp_dir = tempdir().unwrap();
let temp_dir = TempDir::new("content").unwrap();
let config = Arc::new(Config::default());
let content = Content::new(config, temp_dir.path()).await.unwrap();
@ -689,7 +644,7 @@ mod tests {
#[tokio::test]
async fn test_hard_link_task() {
let temp_dir = tempdir().unwrap();
let temp_dir = TempDir::new("content").unwrap();
let config = Arc::new(Config::default());
let content = Content::new(config, temp_dir.path()).await.unwrap();
@ -707,7 +662,7 @@ mod tests {
#[tokio::test]
async fn test_copy_task() {
let temp_dir = tempdir().unwrap();
let temp_dir = TempDir::new("content").unwrap();
let config = Arc::new(Config::default());
let content = Content::new(config, temp_dir.path()).await.unwrap();
@ -723,7 +678,7 @@ mod tests {
#[tokio::test]
async fn test_delete_task() {
let temp_dir = tempdir().unwrap();
let temp_dir = TempDir::new("content").unwrap();
let config = Arc::new(Config::default());
let content = Content::new(config, temp_dir.path()).await.unwrap();
@ -737,7 +692,7 @@ mod tests {
#[tokio::test]
async fn test_read_piece() {
let temp_dir = tempdir().unwrap();
let temp_dir = TempDir::new("content").unwrap();
let config = Arc::new(Config::default());
let content = Content::new(config, temp_dir.path()).await.unwrap();
@ -746,10 +701,7 @@ mod tests {
let data = b"hello, world!";
let mut reader = Cursor::new(data);
content
.write_piece(task_id, 0, 13, &mut reader)
.await
.unwrap();
content.write_piece(task_id, 0, &mut reader).await.unwrap();
let mut reader = content.read_piece(task_id, 0, 13, None).await.unwrap();
let mut buffer = Vec::new();
@ -775,7 +727,7 @@ mod tests {
#[tokio::test]
async fn test_write_piece() {
let temp_dir = tempdir().unwrap();
let temp_dir = TempDir::new("content").unwrap();
let config = Arc::new(Config::default());
let content = Content::new(config, temp_dir.path()).await.unwrap();
@ -784,17 +736,14 @@ mod tests {
let data = b"test";
let mut reader = Cursor::new(data);
let response = content
.write_piece(task_id, 0, 4, &mut reader)
.await
.unwrap();
let response = content.write_piece(task_id, 0, &mut reader).await.unwrap();
assert_eq!(response.length, 4);
assert!(!response.hash.is_empty());
}
#[tokio::test]
async fn test_create_persistent_task() {
let temp_dir = tempdir().unwrap();
let temp_dir = TempDir::new("content").unwrap();
let config = Arc::new(Config::default());
let content = Content::new(config, temp_dir.path()).await.unwrap();
@ -815,7 +764,7 @@ mod tests {
#[tokio::test]
async fn test_hard_link_persistent_cache_task() {
let temp_dir = tempdir().unwrap();
let temp_dir = TempDir::new("content").unwrap();
let config = Arc::new(Config::default());
let content = Content::new(config, temp_dir.path()).await.unwrap();
@ -842,7 +791,7 @@ mod tests {
#[tokio::test]
async fn test_copy_persistent_cache_task() {
let temp_dir = tempdir().unwrap();
let temp_dir = TempDir::new("content").unwrap();
let config = Arc::new(Config::default());
let content = Content::new(config, temp_dir.path()).await.unwrap();
@ -864,7 +813,7 @@ mod tests {
#[tokio::test]
async fn test_delete_persistent_cache_task() {
let temp_dir = tempdir().unwrap();
let temp_dir = TempDir::new("content").unwrap();
let config = Arc::new(Config::default());
let content = Content::new(config, temp_dir.path()).await.unwrap();
@ -881,7 +830,7 @@ mod tests {
#[tokio::test]
async fn test_read_persistent_cache_piece() {
let temp_dir = tempdir().unwrap();
let temp_dir = TempDir::new("content").unwrap();
let config = Arc::new(Config::default());
let content = Content::new(config, temp_dir.path()).await.unwrap();
@ -894,7 +843,7 @@ mod tests {
let data = b"hello, world!";
let mut reader = Cursor::new(data);
content
.write_persistent_cache_piece(task_id, 0, 13, &mut reader)
.write_persistent_cache_piece(task_id, 0, &mut reader)
.await
.unwrap();
@ -925,7 +874,7 @@ mod tests {
#[tokio::test]
async fn test_write_persistent_cache_piece() {
let temp_dir = tempdir().unwrap();
let temp_dir = TempDir::new("content").unwrap();
let config = Arc::new(Config::default());
let content = Content::new(config, temp_dir.path()).await.unwrap();
@ -938,7 +887,7 @@ mod tests {
let data = b"test";
let mut reader = Cursor::new(data);
let response = content
.write_persistent_cache_piece(task_id, 0, 4, &mut reader)
.write_persistent_cache_piece(task_id, 0, &mut reader)
.await
.unwrap();
assert_eq!(response.length, 4);
@ -948,36 +897,14 @@ mod tests {
#[tokio::test]
async fn test_has_enough_space() {
let config = Arc::new(Config::default());
let temp_dir = tempdir().unwrap();
let content = Content::new(config, temp_dir.path()).await.unwrap();
let dir = PathBuf::from("/tmp/dragonfly_test");
let content = Content::new(config, &dir).await.unwrap();
let has_space = content.has_enough_space(1).unwrap();
assert!(has_space);
let has_space = content.has_enough_space(u64::MAX).unwrap();
assert!(!has_space);
let mut config = Config::default();
config.gc.policy.dist_threshold = ByteSize::mib(10);
let config = Arc::new(config);
let content = Content::new(config, temp_dir.path()).await.unwrap();
let file_path = Path::new(temp_dir.path())
.join(DEFAULT_CONTENT_DIR)
.join(DEFAULT_TASK_DIR)
.join("1mib");
let mut file = File::create(&file_path).await.unwrap();
let buffer = vec![0u8; ByteSize::mib(1).as_u64() as usize];
file.write_all(&buffer).await.unwrap();
file.flush().await.unwrap();
let has_space = content
.has_enough_space(ByteSize::mib(9).as_u64() + 1)
.unwrap();
assert!(!has_space);
let has_space = content.has_enough_space(ByteSize::mib(9).as_u64()).unwrap();
assert!(has_space);
}
#[tokio::test]

View File

@ -27,6 +27,7 @@ use std::time::Duration;
use tokio::io::AsyncRead;
use tokio::time::sleep;
use tokio_util::either::Either;
use tokio_util::io::InspectReader;
use tracing::{debug, error, info, instrument, warn};
pub mod cache;
@ -116,15 +117,23 @@ impl Storage {
piece_length: u64,
content_length: u64,
response_header: Option<HeaderMap>,
load_to_cache: bool,
) -> Result<metadata::Task> {
self.content.create_task(id, content_length).await?;
self.metadata.download_task_started(
let metadata = self.metadata.download_task_started(
id,
Some(piece_length),
Some(content_length),
response_header,
)
)?;
self.content.create_task(id, content_length).await?;
if load_to_cache {
let mut cache = self.cache.clone();
cache.put_task(id, content_length).await;
debug!("put task to cache: {}", id);
}
Ok(metadata)
}
/// download_task_finished updates the metadata of the task when the task downloads finished.
@ -376,7 +385,7 @@ impl Storage {
) -> Result<metadata::Piece> {
let response = self
.content
.write_persistent_cache_piece(task_id, offset, length, reader)
.write_persistent_cache_piece(task_id, offset, reader)
.await?;
let digest = Digest::new(Algorithm::Crc32, response.hash);
@ -415,10 +424,11 @@ impl Storage {
offset: u64,
length: u64,
reader: &mut R,
load_to_cache: bool,
timeout: Duration,
) -> Result<metadata::Piece> {
tokio::select! {
piece = self.handle_downloaded_from_source_finished(piece_id, task_id, offset, length, reader) => {
piece = self.handle_downloaded_from_source_finished(piece_id, task_id, offset, length, reader, load_to_cache) => {
piece
}
_ = sleep(timeout) => {
@ -436,11 +446,25 @@ impl Storage {
offset: u64,
length: u64,
reader: &mut R,
load_to_cache: bool,
) -> Result<metadata::Piece> {
let response = self
.content
.write_piece(task_id, offset, length, reader)
.await?;
let response = if load_to_cache {
let mut buffer = Vec::with_capacity(length as usize);
let mut tee = InspectReader::new(reader, |bytes| {
buffer.extend_from_slice(bytes);
});
let response = self.content.write_piece(task_id, offset, &mut tee).await?;
self.cache
.write_piece(task_id, piece_id, bytes::Bytes::from(buffer))
.await?;
debug!("put piece to cache: {}", piece_id);
response
} else {
self.content.write_piece(task_id, offset, reader).await?
};
let digest = Digest::new(Algorithm::Crc32, response.hash);
self.metadata.download_piece_finished(
@ -464,10 +488,11 @@ impl Storage {
expected_digest: &str,
parent_id: &str,
reader: &mut R,
load_to_cache: bool,
timeout: Duration,
) -> Result<metadata::Piece> {
tokio::select! {
piece = self.handle_downloaded_piece_from_parent_finished(piece_id, task_id, offset, length, expected_digest, parent_id, reader) => {
piece = self.handle_downloaded_piece_from_parent_finished(piece_id, task_id, offset, length, expected_digest, parent_id, reader, load_to_cache) => {
piece
}
_ = sleep(timeout) => {
@ -488,11 +513,25 @@ impl Storage {
expected_digest: &str,
parent_id: &str,
reader: &mut R,
load_to_cache: bool,
) -> Result<metadata::Piece> {
let response = self
.content
.write_piece(task_id, offset, length, reader)
.await?;
let response = if load_to_cache {
let mut buffer = Vec::with_capacity(length as usize);
let mut tee = InspectReader::new(reader, |bytes| {
buffer.extend_from_slice(bytes);
});
let response = self.content.write_piece(task_id, offset, &mut tee).await?;
self.cache
.write_piece(task_id, piece_id, bytes::Bytes::from(buffer))
.await?;
debug!("put piece to cache: {}", piece_id);
response
} else {
self.content.write_piece(task_id, offset, reader).await?
};
let length = response.length;
let digest = Digest::new(Algorithm::Crc32, response.hash);
@ -629,7 +668,6 @@ impl Storage {
}
/// download_persistent_cache_piece_from_parent_finished is used for downloading persistent cache piece from parent.
#[allow(clippy::too_many_arguments)]
#[instrument(skip_all)]
pub async fn download_persistent_cache_piece_from_parent_finished<
R: AsyncRead + Unpin + ?Sized,
@ -638,14 +676,13 @@ impl Storage {
piece_id: &str,
task_id: &str,
offset: u64,
length: u64,
expected_digest: &str,
parent_id: &str,
reader: &mut R,
) -> Result<metadata::Piece> {
let response = self
.content
.write_persistent_cache_piece(task_id, offset, length, reader)
.write_persistent_cache_piece(task_id, offset, reader)
.await?;
let length = response.length;

View File

@ -938,7 +938,7 @@ impl Metadata<RocksdbStorageEngine> {
#[cfg(test)]
mod tests {
use super::*;
use tempfile::tempdir;
use tempdir::TempDir;
#[test]
fn test_calculate_digest() {
@ -956,7 +956,7 @@ mod tests {
#[test]
fn should_create_metadata() {
let dir = tempdir().unwrap();
let dir = TempDir::new("metadata").unwrap();
let log_dir = dir.path().join("log");
let metadata = Metadata::new(Arc::new(Config::default()), dir.path(), &log_dir).unwrap();
assert!(metadata.get_tasks().unwrap().is_empty());
@ -968,7 +968,7 @@ mod tests {
#[test]
fn test_task_lifecycle() {
let dir = tempdir().unwrap();
let dir = TempDir::new("metadata").unwrap();
let log_dir = dir.path().join("log");
let metadata = Metadata::new(Arc::new(Config::default()), dir.path(), &log_dir).unwrap();
let task_id = "d3c4e940ad06c47fc36ac67801e6f8e36cb400e2391708620bc7e865b102062c";
@ -1028,7 +1028,7 @@ mod tests {
#[test]
fn test_piece_lifecycle() {
let dir = tempdir().unwrap();
let dir = TempDir::new("metadata").unwrap();
let log_dir = dir.path().join("log");
let metadata = Metadata::new(Arc::new(Config::default()), dir.path(), &log_dir).unwrap();
let task_id = "d3c4e940ad06c47fc36ac67801e6f8e36cb400e2391708620bc7e865b102062c";

View File

@ -252,7 +252,7 @@ where
mod tests {
use super::*;
use serde::{Deserialize, Serialize};
use tempfile::tempdir;
use tempdir::TempDir;
#[derive(Debug, Serialize, Deserialize, PartialEq, Clone)]
struct Object {
@ -265,7 +265,7 @@ mod tests {
}
fn create_test_engine() -> RocksdbStorageEngine {
let temp_dir = tempdir().unwrap();
let temp_dir = TempDir::new("rocksdb_test").unwrap();
let log_dir = temp_dir.path().to_path_buf();
RocksdbStorageEngine::open(temp_dir.path(), &log_dir, &[Object::NAMESPACE], false).unwrap()
}

View File

@ -13,6 +13,7 @@ edition.workspace = true
dragonfly-client-core.workspace = true
dragonfly-api.workspace = true
reqwest.workspace = true
hyper.workspace = true
http-range-header.workspace = true
http.workspace = true
tracing.workspace = true
@ -23,15 +24,14 @@ rustls-pki-types.workspace = true
rustls-pemfile.workspace = true
sha2.workspace = true
uuid.workspace = true
sysinfo.workspace = true
hex.workspace = true
crc32fast.workspace = true
openssl.workspace = true
crc32fast.workspace = true
lazy_static.workspace = true
bytesize.workspace = true
lru.workspace = true
tokio.workspace = true
rustix = { version = "1.0.8", features = ["fs"] }
rustix = { version = "1.0.5", features = ["fs"] }
base64 = "0.22.1"
pnet = "0.35.0"

View File

@ -14,7 +14,7 @@
* limitations under the License.
*/
use dragonfly_client_core::{Error as ClientError, Result as ClientResult};
use dragonfly_client_core::Result as ClientResult;
use sha2::Digest as Sha2Digest;
use std::fmt;
use std::io::{self, Read};
@ -112,36 +112,9 @@ impl FromStr for Digest {
}
let algorithm = match parts[0] {
"crc32" => {
if parts[1].len() != 10 {
return Err(format!(
"invalid crc32 digest length: {}, expected 10",
parts[1].len()
));
}
Algorithm::Crc32
}
"sha256" => {
if parts[1].len() != 64 {
return Err(format!(
"invalid sha256 digest length: {}, expected 64",
parts[1].len()
));
}
Algorithm::Sha256
}
"sha512" => {
if parts[1].len() != 128 {
return Err(format!(
"invalid sha512 digest length: {}, expected 128",
parts[1].len()
));
}
Algorithm::Sha512
}
"crc32" => Algorithm::Crc32,
"sha256" => Algorithm::Sha256,
"sha512" => Algorithm::Sha512,
_ => return Err(format!("invalid digest algorithm: {}", parts[0])),
};
@ -182,25 +155,6 @@ pub fn calculate_file_digest(algorithm: Algorithm, path: &Path) -> ClientResult<
}
}
/// verify_file_digest verifies the digest of a file against an expected digest.
pub fn verify_file_digest(expected_digest: Digest, file_path: &Path) -> ClientResult<()> {
let digest = match calculate_file_digest(expected_digest.algorithm(), file_path) {
Ok(digest) => digest,
Err(err) => {
return Err(err);
}
};
if digest.to_string() != expected_digest.to_string() {
return Err(ClientError::DigestMismatch(
expected_digest.to_string(),
digest.to_string(),
));
}
Ok(())
}
#[cfg(test)]
mod tests {
use super::*;
@ -251,28 +205,4 @@ mod tests {
calculate_file_digest(Algorithm::Crc32, path).expect("failed to calculate Crc32 hash");
assert_eq!(digest.encoded(), expected_crc32);
}
#[test]
fn test_verify_file_digest() {
let content = b"test content";
let temp_file = tempfile::NamedTempFile::new().expect("failed to create temp file");
let path = temp_file.path();
let mut file = File::create(path).expect("failed to create file");
file.write_all(content).expect("failed to write to file");
let expected_sha256_digest = Digest::new(
Algorithm::Sha256,
"6ae8a75555209fd6c44157c0aed8016e763ff435a19cf186f76863140143ff72".to_string(),
);
assert!(verify_file_digest(expected_sha256_digest, path).is_ok());
let expected_sha512_digest = Digest::new(
Algorithm::Sha512,
"0cbf4caef38047bba9a24e621a961484e5d2a92176a859e7eb27df343dd34eb98d538a6c5f4da1ce302ec250b821cc001e46cc97a704988297185a4df7e99602".to_string(),
);
assert!(verify_file_digest(expected_sha512_digest, path).is_ok());
let expected_crc32_digest = Digest::new(Algorithm::Crc32, "1475635037".to_string());
assert!(verify_file_digest(expected_crc32_digest, path).is_ok());
}
}

View File

@ -14,15 +14,48 @@
* limitations under the License.
*/
use bytesize::ByteSize;
use bytesize::{ByteSize, MB};
use pnet::datalink::{self, NetworkInterface};
use std::cmp::min;
use std::net::IpAddr;
use std::sync::Arc;
use std::time::Duration;
use sysinfo::Networks;
use tokio::sync::Mutex;
use tracing::{info, warn};
#[cfg(not(target_os = "linux"))]
use tracing::warn;
/// get_interface_by_ip returns the name of the network interface that has the specified IP
/// address.
pub fn get_interface_by_ip(ip: IpAddr) -> Option<NetworkInterface> {
for interface in datalink::interfaces() {
for ip_network in interface.ips.iter() {
if ip_network.ip() == ip {
return Some(interface);
}
}
}
None
}
/// get_interface_speed_by_ip returns the speed of the network interface that has the specified IP
/// address in Mbps.
pub fn get_interface_speed(interface_name: &str) -> Option<u64> {
#[cfg(target_os = "linux")]
{
let speed_path = format!("/sys/class/net/{}/speed", interface_name);
std::fs::read_to_string(&speed_path)
.ok()
.and_then(|speed_str| speed_str.trim().parse::<u64>().ok())
}
#[cfg(not(target_os = "linux"))]
{
warn!(
"can not get interface {} speed on non-linux platform",
interface_name
);
None
}
}
/// Interface represents a network interface with its information.
#[derive(Debug, Clone, Default)]
@ -30,201 +63,23 @@ pub struct Interface {
/// name is the name of the network interface.
pub name: String,
/// bandwidth is the bandwidth of the network interface in bps.
// bandwidth is the bandwidth of the network interface in Mbps.
pub bandwidth: u64,
// network_data_mutex is a mutex to protect access to network data.
network_data_mutex: Arc<Mutex<()>>,
}
/// NetworkData represents the network data for a specific interface,
#[derive(Debug, Clone, Default)]
pub struct NetworkData {
/// max_rx_bandwidth is the maximum receive bandwidth of the interface in bps.
pub max_rx_bandwidth: u64,
/// get_interface_info returns the network interface information for the specified IP address.
pub fn get_interface_info(ip: IpAddr, rate_limit: ByteSize) -> Option<Interface> {
let rate_limit = rate_limit.as_u64() / MB * 8; // convert to Mbps
/// rx_bandwidth is the current receive bandwidth of the interface in bps.
pub rx_bandwidth: Option<u64>,
/// max_tx_bandwidth is the maximum transmit bandwidth of the interface in bps.
pub max_tx_bandwidth: u64,
/// tx_bandwidth is the current transmit bandwidth of the interface in bps.
pub tx_bandwidth: Option<u64>,
}
/// Interface methods provide functionality to get network interface information.
impl Interface {
/// DEFAULT_NETWORKS_REFRESH_INTERVAL is the default interval for refreshing network data.
const DEFAULT_NETWORKS_REFRESH_INTERVAL: Duration = Duration::from_secs(2);
/// new creates a new Interface instance based on the provided IP address and rate limit.
pub fn new(ip: IpAddr, rate_limit: ByteSize) -> Interface {
let rate_limit = Self::byte_size_to_bits(rate_limit); // convert to bps
let Some(interface) = Self::get_network_interface_by_ip(ip) else {
warn!(
"can not find interface for IP address {}, network interface unknown with bandwidth {} bps",
ip, rate_limit
);
return Interface {
name: "unknown".to_string(),
bandwidth: rate_limit,
network_data_mutex: Arc::new(Mutex::new(())),
};
};
match Self::get_speed(&interface.name) {
Some(speed) => {
let bandwidth = min(Self::megabits_to_bits(speed), rate_limit);
info!(
"network interface {} with bandwidth {} bps",
interface.name, bandwidth
);
Interface {
name: interface.name,
bandwidth,
network_data_mutex: Arc::new(Mutex::new(())),
}
}
None => {
warn!(
"can not get speed, network interface {} with bandwidth {} bps",
interface.name, rate_limit
);
Interface {
name: interface.name,
bandwidth: rate_limit,
network_data_mutex: Arc::new(Mutex::new(())),
}
}
}
}
/// get_network_data retrieves the network data for the interface.
pub async fn get_network_data(&self) -> NetworkData {
// Lock the mutex to ensure exclusive access to network data.
let _guard = self.network_data_mutex.lock().await;
// Initialize sysinfo network.
let mut networks = Networks::new_with_refreshed_list();
// Sleep to calculate the network traffic difference over
// the DEFAULT_NETWORKS_REFRESH_INTERVAL.
tokio::time::sleep(Self::DEFAULT_NETWORKS_REFRESH_INTERVAL).await;
// Refresh network information.
networks.refresh();
let Some(network_data) = networks.get(self.name.as_str()) else {
warn!("can not find network data for interface {}", self.name);
return NetworkData {
max_rx_bandwidth: self.bandwidth,
max_tx_bandwidth: self.bandwidth,
..Default::default()
};
};
// Calculate the receive and transmit bandwidth in bits per second.
let rx_bandwidth = (Self::bytes_to_bits(network_data.received()) as f64
/ Self::DEFAULT_NETWORKS_REFRESH_INTERVAL.as_secs_f64())
.round() as u64;
// Calculate the transmit bandwidth in bits per second.
let tx_bandwidth = (Self::bytes_to_bits(network_data.transmitted()) as f64
/ Self::DEFAULT_NETWORKS_REFRESH_INTERVAL.as_secs_f64())
.round() as u64;
NetworkData {
max_rx_bandwidth: self.bandwidth,
rx_bandwidth: Some(rx_bandwidth),
max_tx_bandwidth: self.bandwidth,
tx_bandwidth: Some(tx_bandwidth),
}
}
/// get_speed returns the speed of the network interface in Mbps.
pub fn get_speed(name: &str) -> Option<u64> {
#[cfg(target_os = "linux")]
{
let speed_path = format!("/sys/class/net/{}/speed", name);
std::fs::read_to_string(&speed_path)
.ok()
.and_then(|speed_str| speed_str.trim().parse::<u64>().ok())
}
#[cfg(not(target_os = "linux"))]
{
warn!("can not get interface {} speed on non-linux platform", name);
None
}
}
/// get_network_interface_by_ip returns the network interface that has the specified
/// IP address.
pub fn get_network_interface_by_ip(ip: IpAddr) -> Option<NetworkInterface> {
datalink::interfaces()
.into_iter()
.find(|interface| interface.ips.iter().any(|ip_net| ip_net.ip() == ip))
}
/// byte_size_to_bits converts a ByteSize to bits.
pub fn byte_size_to_bits(size: ByteSize) -> u64 {
size.as_u64() * 8
}
/// megabits_to_bit converts megabits to bits.
pub fn megabits_to_bits(size: u64) -> u64 {
size * 1_000_000 // 1 Mbit = 1,000,000 bits
}
/// bytes_to_bits converts bytes to bits.
pub fn bytes_to_bits(size: u64) -> u64 {
size * 8 // 1 byte = 8 bits
}
}
#[cfg(test)]
mod tests {
use super::*;
use bytesize::ByteSize;
#[test]
fn test_byte_size_to_bits() {
let test_cases = vec![
(ByteSize::kb(1), 8_000u64),
(ByteSize::mb(1), 8_000_000u64),
(ByteSize::gb(1), 8_000_000_000u64),
(ByteSize::b(0), 0u64),
];
for (input, expected) in test_cases {
let result = Interface::byte_size_to_bits(input);
assert_eq!(result, expected);
}
}
#[test]
fn test_megabits_to_bits() {
let test_cases = vec![
(1u64, 1_000_000u64),
(1000u64, 1_000_000_000u64),
(0u64, 0u64),
];
for (input, expected) in test_cases {
let result = Interface::megabits_to_bits(input);
assert_eq!(result, expected);
}
}
#[test]
fn test_bytes_to_bits() {
let test_cases = vec![(1u64, 8u64), (1000u64, 8_000u64), (0u64, 0u64)];
for (input, expected) in test_cases {
let result = Interface::bytes_to_bits(input);
assert_eq!(result, expected);
}
let interface = get_interface_by_ip(ip)?;
match get_interface_speed(&interface.name) {
Some(speed) => Some(Interface {
name: interface.name,
bandwidth: min(speed, rate_limit),
}),
None => Some(Interface {
name: interface.name,
bandwidth: rate_limit,
}),
}
}

View File

@ -34,6 +34,8 @@ hyper.workspace = true
hyper-util.workspace = true
hyper-rustls.workspace = true
tracing.workspace = true
validator.workspace = true
humantime.workspace = true
serde.workspace = true
chrono.workspace = true
prost-wkt-types.workspace = true
@ -53,40 +55,35 @@ clap.workspace = true
anyhow.workspace = true
bytes.workspace = true
bytesize.workspace = true
humantime.workspace = true
uuid.workspace = true
percent-encoding.workspace = true
tokio-rustls.workspace = true
serde_json.workspace = true
lru.workspace = true
fs2.workspace = true
lazy_static.workspace = true
futures.workspace = true
local-ip-address.workspace = true
sysinfo.workspace = true
tracing-appender = "0.2.3"
tracing-subscriber = { version = "0.3", features = ["env-filter", "time", "chrono"] }
tracing-panic = "0.1.2"
tracing-opentelemetry = "0.30.0"
opentelemetry = { version = "0.29.1", default-features = false, features = ["trace"] }
opentelemetry-otlp = { version = "0.29.0", default-features = false, features = ["trace", "grpc-tonic", "http-proto", "reqwest-blocking-client"] }
opentelemetry_sdk = { version = "0.29.0", default-features = false, features = ["trace", "rt-tokio"] }
opentelemetry-semantic-conventions = { version = "0.30.0", features = ["semconv_experimental"] }
opentelemetry-otlp = { version = "0.29.0", default-features = false, features = ["trace", "grpc-tonic"] }
opentelemetry_sdk = { version = "0.29.0", default-features = false, features = ["trace"] }
opentelemetry-semantic-conventions = { version = "0.29.0", features = ["semconv_experimental"] }
rolling-file = "0.2.0"
pprof = { version = "0.15", features = ["flamegraph", "protobuf-codec"] }
pprof = { version = "0.14", features = ["flamegraph", "protobuf-codec"] }
prometheus = { version = "0.13", features = ["process"] }
tonic-health = "0.12.3"
sysinfo = { version = "0.32.1", default-features = false, features = ["component", "disk", "network", "system", "user"] }
tower = { version = "0.4.13", features = ["limit", "load-shed", "buffer"] }
indicatif = "0.18.0"
indicatif = "0.17.11"
hashring = "0.3.6"
fslock = "0.2.1"
leaky-bucket = "1.1.2"
http-body-util = "0.1.3"
termion = "4.0.5"
tabled = "0.20.0"
tabled = "0.19.0"
path-absolutize = "3.1.1"
dashmap = "6.1.0"
fastrand = "2.3.0"
glob = "0.3.3"
console-subscriber = "0.4.1"
[dev-dependencies]
tempfile.workspace = true

View File

@ -14,9 +14,10 @@
* limitations under the License.
*/
use crate::grpc::scheduler::SchedulerClient;
use crate::grpc::{manager::ManagerClient, scheduler::SchedulerClient};
use crate::shutdown;
use dragonfly_api::common::v2::{Build, Cpu, Disk, Host, Memory, Network};
use dragonfly_api::manager::v2::{DeleteSeedPeerRequest, SourceType, UpdateSeedPeerRequest};
use dragonfly_api::scheduler::v2::{AnnounceHostRequest, DeleteHostRequest};
use dragonfly_client_config::{
dfdaemon::{Config, HostType},
@ -24,13 +25,89 @@ use dragonfly_client_config::{
};
use dragonfly_client_core::error::{ErrorType, OrErr};
use dragonfly_client_core::Result;
use dragonfly_client_util::net::Interface;
use std::env;
use std::sync::Arc;
use std::time::Duration;
use sysinfo::System;
use tokio::sync::mpsc;
use tracing::{debug, error, info, instrument};
use tracing::{error, info, instrument};
/// ManagerAnnouncer is used to announce the dfdaemon information to the manager.
pub struct ManagerAnnouncer {
/// config is the configuration of the dfdaemon.
config: Arc<Config>,
/// manager_client is the grpc client of the manager.
manager_client: Arc<ManagerClient>,
/// shutdown is used to shutdown the announcer.
shutdown: shutdown::Shutdown,
/// _shutdown_complete is used to notify the announcer is shutdown.
_shutdown_complete: mpsc::UnboundedSender<()>,
}
/// ManagerAnnouncer implements the manager announcer of the dfdaemon.
impl ManagerAnnouncer {
/// new creates a new manager announcer.
pub fn new(
config: Arc<Config>,
manager_client: Arc<ManagerClient>,
shutdown: shutdown::Shutdown,
shutdown_complete_tx: mpsc::UnboundedSender<()>,
) -> Self {
Self {
config,
manager_client,
shutdown,
_shutdown_complete: shutdown_complete_tx,
}
}
/// run announces the dfdaemon information to the manager.
pub async fn run(&self) -> Result<()> {
// Clone the shutdown channel.
let mut shutdown = self.shutdown.clone();
// If the seed peer is enabled, we should announce the seed peer to the manager.
if self.config.seed_peer.enable {
// Register the seed peer to the manager.
self.manager_client
.update_seed_peer(UpdateSeedPeerRequest {
source_type: SourceType::SeedPeerSource.into(),
hostname: self.config.host.hostname.clone(),
r#type: self.config.seed_peer.kind.to_string(),
idc: self.config.host.idc.clone(),
location: self.config.host.location.clone(),
ip: self.config.host.ip.unwrap().to_string(),
port: self.config.upload.server.port as i32,
download_port: self.config.upload.server.port as i32,
seed_peer_cluster_id: self.config.seed_peer.cluster_id,
})
.await?;
// Announce to scheduler shutting down with signals.
shutdown.recv().await;
// Delete the seed peer from the manager.
self.manager_client
.delete_seed_peer(DeleteSeedPeerRequest {
source_type: SourceType::SeedPeerSource.into(),
hostname: self.config.host.hostname.clone(),
ip: self.config.host.ip.unwrap().to_string(),
seed_peer_cluster_id: self.config.seed_peer.cluster_id,
})
.await?;
info!("announce to manager shutting down");
} else {
shutdown.recv().await;
info!("announce to manager shutting down");
}
Ok(())
}
}
/// Announcer is used to announce the dfdaemon information to the manager and scheduler.
pub struct SchedulerAnnouncer {
@ -43,9 +120,6 @@ pub struct SchedulerAnnouncer {
/// scheduler_client is the grpc client of the scheduler.
scheduler_client: Arc<SchedulerClient>,
/// interface is the network interface.
interface: Arc<Interface>,
/// shutdown is used to shutdown the announcer.
shutdown: shutdown::Shutdown,
@ -60,7 +134,6 @@ impl SchedulerAnnouncer {
config: Arc<Config>,
host_id: String,
scheduler_client: Arc<SchedulerClient>,
interface: Arc<Interface>,
shutdown: shutdown::Shutdown,
shutdown_complete_tx: mpsc::UnboundedSender<()>,
) -> Result<Self> {
@ -68,7 +141,6 @@ impl SchedulerAnnouncer {
config,
host_id,
scheduler_client,
interface,
shutdown,
_shutdown_complete: shutdown_complete_tx,
};
@ -76,7 +148,7 @@ impl SchedulerAnnouncer {
// Initialize the scheduler announcer.
announcer
.scheduler_client
.init_announce_host(announcer.make_announce_host_request(Duration::ZERO).await?)
.init_announce_host(announcer.make_announce_host_request(Duration::ZERO)?)
.await?;
Ok(announcer)
}
@ -91,7 +163,7 @@ impl SchedulerAnnouncer {
loop {
tokio::select! {
_ = interval.tick() => {
let request = match self.make_announce_host_request(interval.period()).await {
let request = match self.make_announce_host_request(interval.period()) {
Ok(request) => request,
Err(err) => {
error!("make announce host request failed: {}", err);
@ -120,7 +192,7 @@ impl SchedulerAnnouncer {
/// make_announce_host_request makes the announce host request.
#[instrument(skip_all)]
async fn make_announce_host_request(&self, interval: Duration) -> Result<AnnounceHostRequest> {
fn make_announce_host_request(&self, interval: Duration) -> Result<AnnounceHostRequest> {
// If the seed peer is enabled, we should announce the seed peer to the scheduler.
let host_type = if self.config.seed_peer.enable {
self.config.seed_peer.kind
@ -156,25 +228,25 @@ impl SchedulerAnnouncer {
free: sys.free_memory(),
};
// Wait for getting the network data.
let network_data = self.interface.get_network_data().await;
debug!(
"network data: rx bandwidth {}/{} bps, tx bandwidth {}/{} bps",
network_data.rx_bandwidth.unwrap_or(0),
network_data.max_rx_bandwidth,
network_data.tx_bandwidth.unwrap_or(0),
network_data.max_tx_bandwidth
);
// Get the network information.
let network = Network {
// TODO: Get the count of the tcp connection.
tcp_connection_count: 0,
// TODO: Get the count of the upload tcp connection.
upload_tcp_connection_count: 0,
idc: self.config.host.idc.clone(),
location: self.config.host.location.clone(),
max_rx_bandwidth: network_data.max_rx_bandwidth,
rx_bandwidth: network_data.rx_bandwidth,
max_tx_bandwidth: network_data.max_tx_bandwidth,
tx_bandwidth: network_data.tx_bandwidth,
..Default::default()
// TODO: Get the network download rate, refer to
// https://docs.rs/sysinfo/latest/sysinfo/struct.NetworkData.html#method.received.
download_rate: 0,
download_rate_limit: self.config.download.rate_limit.as_u64(),
// TODO: Get the network download rate, refer to
// https://docs.rs/sysinfo/latest/sysinfo/struct.NetworkData.html#method.transmitted
upload_rate: 0,
upload_rate_limit: self.config.upload.rate_limit.as_u64(),
};
// Get the disk information.

View File

@ -25,7 +25,6 @@ use dragonfly_client_core::{
};
use dragonfly_client_util::fs::fallocate;
use indicatif::{ProgressBar, ProgressState, ProgressStyle};
use local_ip_address::local_ip;
use path_absolutize::*;
use std::path::{Path, PathBuf};
use std::time::Duration;
@ -86,13 +85,6 @@ pub struct ExportCommand {
)]
timeout: Duration,
#[arg(
long = "digest",
required = false,
help = "Verify the integrity of the downloaded file using the specified digest, support sha256, sha512, crc32. If the digest is not specified, the downloaded file will not be verified. Format: <algorithm>:<digest>, e.g. sha256:1234567890abcdef1234567890abcdef1234567890abcdef1234567890abcdef, crc32:12345678"
)]
digest: Option<String>,
#[arg(
short = 'e',
long = "endpoint",
@ -129,13 +121,7 @@ pub struct ExportCommand {
/// Implement the execute for ExportCommand.
impl ExportCommand {
/// Executes the export command with comprehensive validation and advanced error handling.
///
/// This function serves as the main entry point for the dfcache export command execution.
/// It handles the complete workflow including argument parsing, validation, logging setup,
/// dfdaemon client connection, and export operation execution. The function provides
/// sophisticated error reporting with colored terminal output, including specialized
/// handling for backend errors with HTTP status codes and headers.
/// execute executes the export command.
pub async fn execute(&self) -> Result<()> {
// Parse command line arguments.
Args::parse();
@ -148,9 +134,6 @@ impl ExportCommand {
self.log_max_files,
None,
None,
None,
None,
None,
false,
self.console,
);
@ -442,13 +425,7 @@ impl ExportCommand {
Ok(())
}
/// Executes the export operation to retrieve cached files from the persistent cache system.
///
/// This function handles the core export functionality by downloading a cached file from the
/// dfdaemon persistent cache system. It supports two transfer modes: direct file transfer
/// by dfdaemon (hardlink/copy) or streaming piece content through the client for manual
/// file assembly. The operation provides real-time progress feedback and handles file
/// creation, directory setup, and efficient piece-by-piece writing with sparse file allocation.
/// run runs the export command.
async fn run(&self, dfdaemon_download_client: DfdaemonDownloadClient) -> Result<()> {
// Dfcache needs to notify dfdaemon to transfer the piece content of downloading file via unix domain socket
// when the `transfer_from_dfdaemon` is true. Otherwise, dfdaemon will download the file and hardlink or
@ -477,8 +454,6 @@ impl ExportCommand {
),
need_piece_content,
force_hard_link: self.force_hard_link,
digest: self.digest.clone(),
remote_ip: Some(local_ip().unwrap().to_string()),
})
.await
.inspect_err(|err| {
@ -577,12 +552,7 @@ impl ExportCommand {
Ok(())
}
/// Validates command line arguments for the export operation to ensure safe file output.
///
/// This function performs essential validation of the output path to prevent file conflicts
/// and ensure the target location is suitable for export operations. It checks parent
/// directory existence, prevents accidental file overwrites, and validates path accessibility
/// before allowing the export operation to proceed.
/// validate_args validates the command line arguments.
fn validate_args(&self) -> Result<()> {
let absolute_path = Path::new(&self.output).absolutize()?;
match absolute_path.parent() {

View File

@ -24,7 +24,6 @@ use dragonfly_client_core::{
Error, Result,
};
use indicatif::{ProgressBar, ProgressStyle};
use local_ip_address::local_ip;
use path_absolutize::*;
use std::path::{Path, PathBuf};
use std::time::Duration;
@ -128,13 +127,7 @@ pub struct ImportCommand {
/// Implement the execute for ImportCommand.
impl ImportCommand {
/// Executes the import sub command with comprehensive validation and error handling.
///
/// This function serves as the main entry point for the dfcache import command execution.
/// It handles the complete workflow including argument parsing, validation, logging setup,
/// dfdaemon client connection, and import operation execution. The function provides
/// detailed error reporting with colored terminal output and follows a fail-fast approach
/// with immediate process termination on any critical failures.
/// execute executes the import sub command.
pub async fn execute(&self) -> Result<()> {
// Parse command line arguments.
Args::parse();
@ -147,9 +140,6 @@ impl ImportCommand {
self.log_max_files,
None,
None,
None,
None,
None,
false,
self.console,
);
@ -332,13 +322,7 @@ impl ImportCommand {
Ok(())
}
/// Executes the cache import operation by uploading a file to the persistent cache system.
///
/// This function handles the core import functionality by uploading a local file to the
/// dfdaemon persistent cache system. It provides visual feedback through a progress spinner,
/// converts the file path to absolute format, and configures the cache task with specified
/// parameters including TTL, replica count, and piece length. The operation is asynchronous
/// and provides completion feedback with the generated task ID.
/// run runs the import sub command.
async fn run(&self, dfdaemon_download_client: DfdaemonDownloadClient) -> Result<()> {
let absolute_path = Path::new(&self.path).absolutize()?;
info!("import file: {}", absolute_path.to_string_lossy());
@ -367,7 +351,6 @@ impl ImportCommand {
prost_wkt_types::Duration::try_from(self.timeout)
.or_err(ErrorType::ParseError)?,
),
remote_ip: Some(local_ip().unwrap().to_string()),
})
.await?;
@ -375,12 +358,7 @@ impl ImportCommand {
Ok(())
}
/// Validates command line arguments for the import operation to ensure safe and correct execution.
///
/// This function performs comprehensive validation of import-specific parameters to prevent
/// invalid operations and ensure the import request meets all system requirements. It validates
/// TTL boundaries, file existence and type, and piece length constraints before allowing the
/// import operation to proceed.
/// validate_args validates the command line arguments.
fn validate_args(&self) -> Result<()> {
if self.ttl < Duration::from_secs(5 * 60)
|| self.ttl > Duration::from_secs(7 * 24 * 60 * 60)

View File

@ -106,12 +106,7 @@ async fn main() -> anyhow::Result<()> {
Ok(())
}
/// Creates and validates a dfdaemon download client with health checking.
///
/// This function establishes a connection to the dfdaemon service via Unix domain socket
/// and performs a health check to ensure the service is running and ready to handle
/// download requests. Only after successful health verification does it return the
/// download client for actual use.
/// get_and_check_dfdaemon_download_client gets a dfdaemon download client and checks its health.
pub async fn get_dfdaemon_download_client(endpoint: PathBuf) -> Result<DfdaemonDownloadClient> {
// Check dfdaemon's health.
let health_client = HealthClient::new_unix(endpoint.clone()).await?;

View File

@ -22,7 +22,6 @@ use dragonfly_client_core::{
Error, Result,
};
use humantime::format_duration;
use local_ip_address::local_ip;
use std::time::Duration;
use tabled::{
settings::{object::Rows, Alignment, Modify, Style},
@ -74,13 +73,7 @@ pub struct StatCommand {
/// Implement the execute for StatCommand.
impl StatCommand {
/// Executes the stat command with comprehensive error handling and user feedback.
///
/// This function serves as the main entry point for the dfcache stat command execution.
/// It handles the complete lifecycle including argument parsing, logging initialization,
/// dfdaemon client setup, and command execution with detailed error reporting. The
/// function provides colored terminal output for better user experience and exits
/// with appropriate status codes on failure.
/// execute executes the stat command.
pub async fn execute(&self) -> Result<()> {
// Parse command line arguments.
Args::parse();
@ -93,9 +86,6 @@ impl StatCommand {
self.log_max_files,
None,
None,
None,
None,
None,
false,
self.console,
);
@ -240,17 +230,11 @@ impl StatCommand {
Ok(())
}
/// Executes the stat command to retrieve and display persistent cache task information.
///
/// This function queries the dfdaemon service for detailed information about a specific
/// persistent cache task and presents it in a formatted table for user consumption.
/// It handles data conversion from raw protocol buffer values to human-readable formats
/// including byte sizes, durations, and timestamps with proper timezone conversion.
/// run runs the stat command.
async fn run(&self, dfdaemon_download_client: DfdaemonDownloadClient) -> Result<()> {
let task = dfdaemon_download_client
.stat_persistent_cache_task(StatPersistentCacheTaskRequest {
task_id: self.id.clone(),
remote_ip: Some(local_ip().unwrap().to_string()),
})
.await?;

View File

@ -15,7 +15,7 @@
*/
use clap::Parser;
use dragonfly_client::announcer::SchedulerAnnouncer;
use dragonfly_client::announcer::{ManagerAnnouncer, SchedulerAnnouncer};
use dragonfly_client::dynconfig::Dynconfig;
use dragonfly_client::gc::GC;
use dragonfly_client::grpc::{
@ -32,7 +32,7 @@ use dragonfly_client::tracing::init_tracing;
use dragonfly_client_backend::BackendFactory;
use dragonfly_client_config::{dfdaemon, VersionValueParser};
use dragonfly_client_storage::Storage;
use dragonfly_client_util::{id_generator::IDGenerator, net::Interface};
use dragonfly_client_util::id_generator::IDGenerator;
use std::net::SocketAddr;
use std::path::PathBuf;
use std::sync::Arc;
@ -145,10 +145,7 @@ async fn main() -> Result<(), anyhow::Error> {
args.log_dir.clone(),
args.log_level,
args.log_max_files,
config.tracing.protocol.clone(),
config.tracing.endpoint.clone(),
config.tracing.path.clone(),
Some(config.tracing.headers.clone()),
config.tracing.addr.to_owned(),
Some(config.host.clone()),
config.seed_peer.enable,
args.console,
@ -229,9 +226,6 @@ async fn main() -> Result<(), anyhow::Error> {
)?;
let persistent_cache_task = Arc::new(persistent_cache_task);
let interface = Interface::new(config.host.ip.unwrap(), config.upload.rate_limit);
let interface = Arc::new(interface);
// Initialize health server.
let health = Health::new(
SocketAddr::new(config.health.server.ip.unwrap(), config.health.server.port),
@ -261,12 +255,19 @@ async fn main() -> Result<(), anyhow::Error> {
shutdown_complete_tx.clone(),
);
// Initialize manager announcer.
let manager_announcer = ManagerAnnouncer::new(
config.clone(),
manager_client.clone(),
shutdown.clone(),
shutdown_complete_tx.clone(),
);
// Initialize scheduler announcer.
let scheduler_announcer = SchedulerAnnouncer::new(
config.clone(),
id_generator.host_id(),
scheduler_client.clone(),
interface.clone(),
shutdown.clone(),
shutdown_complete_tx.clone(),
)
@ -281,7 +282,6 @@ async fn main() -> Result<(), anyhow::Error> {
SocketAddr::new(config.upload.server.ip.unwrap(), config.upload.server.port),
task.clone(),
persistent_cache_task.clone(),
interface.clone(),
shutdown.clone(),
shutdown_complete_tx.clone(),
);
@ -330,6 +330,10 @@ async fn main() -> Result<(), anyhow::Error> {
info!("stats server exited");
},
_ = tokio::spawn(async move { manager_announcer.run().await.unwrap_or_else(|err| error!("announcer manager failed: {}", err))} ) => {
info!("announcer manager exited");
},
_ = tokio::spawn(async move { scheduler_announcer.run().await }) => {
info!("announcer scheduler exited");
},

View File

@ -17,30 +17,32 @@
use bytesize::ByteSize;
use clap::Parser;
use dragonfly_api::common::v2::{Download, Hdfs, ObjectStorage, TaskType};
use dragonfly_api::dfdaemon::v2::{
download_task_response, DownloadTaskRequest, ListTaskEntriesRequest,
};
use dragonfly_api::dfdaemon::v2::{download_task_response, DownloadTaskRequest};
use dragonfly_api::errordetails::v2::Backend;
use dragonfly_client::grpc::dfdaemon_download::DfdaemonDownloadClient;
use dragonfly_client::grpc::health::HealthClient;
use dragonfly_client::metrics::{
collect_backend_request_failure_metrics, collect_backend_request_finished_metrics,
collect_backend_request_started_metrics,
};
use dragonfly_client::resource::piece::MIN_PIECE_LENGTH;
use dragonfly_client::tracing::init_tracing;
use dragonfly_client_backend::{hdfs, object_storage, BackendFactory, DirEntry};
use dragonfly_client_backend::{hdfs, object_storage, BackendFactory, DirEntry, HeadRequest};
use dragonfly_client_config::VersionValueParser;
use dragonfly_client_config::{self, dfdaemon, dfget};
use dragonfly_client_core::error::{ErrorType, OrErr};
use dragonfly_client_core::error::{BackendError, ErrorType, OrErr};
use dragonfly_client_core::{Error, Result};
use dragonfly_client_util::{fs::fallocate, http::header_vec_to_hashmap};
use glob::Pattern;
use dragonfly_client_util::{
fs::fallocate,
http::{header_vec_to_hashmap, header_vec_to_headermap},
};
use indicatif::{MultiProgress, ProgressBar, ProgressState, ProgressStyle};
use local_ip_address::local_ip;
use path_absolutize::*;
use percent_encoding::percent_decode_str;
use std::collections::{HashMap, HashSet};
use std::path::{Component, Path, PathBuf};
use std::path::{Path, PathBuf};
use std::str::FromStr;
use std::sync::Arc;
use std::time::Duration;
use std::time::{Duration, Instant};
use std::{cmp::min, fmt::Write};
use termion::{color, style};
use tokio::fs::{self, OpenOptions};
@ -111,7 +113,7 @@ struct Args {
#[arg(
long = "content-for-calculating-task-id",
help = "Specify the content used to calculate the task ID. If it is set, use its value to calculate the task ID, Otherwise, calculate the task ID based on URL, piece-length, tag, application, and filtered-query-params."
help = "Specify the content used to calculate the task ID. If it is set, use its value to calculate the task ID, Otherwise, calculate the task ID based on url, piece-length, tag, application, and filtered-query-params."
)]
content_for_calculating_task_id: Option<String>,
@ -139,11 +141,12 @@ struct Args {
timeout: Duration,
#[arg(
short = 'd',
long = "digest",
required = false,
help = "Verify the integrity of the downloaded file using the specified digest, support sha256, sha512, crc32. If the digest is not specified, the downloaded file will not be verified. Format: <algorithm>:<digest>. Examples: sha256:1234567890abcdef1234567890abcdef1234567890abcdef1234567890abcdef, crc32:12345678"
default_value = "",
help = "Verify the integrity of the downloaded file using the specified digest, e.g. md5:86d3f3a95c324c9479bd8986968f4327"
)]
digest: Option<String>,
digest: String,
#[arg(
short = 'p',
@ -163,14 +166,14 @@ struct Args {
#[arg(
long = "application",
default_value = "",
help = "Different applications for the same URL will be divided into different tasks"
help = "Different applications for the same url will be divided into different tasks"
)]
application: String,
#[arg(
long = "tag",
default_value = "",
help = "Different tags for the same URL will be divided into different tasks"
help = "Different tags for the same url will be divided into different tasks"
)]
tag: String,
@ -178,24 +181,17 @@ struct Args {
short = 'H',
long = "header",
required = false,
help = "Specify the header for downloading file. Examples: --header='Content-Type: application/json' --header='Accept: application/json'"
help = "Specify the header for downloading file, e.g. --header='Content-Type: application/json' --header='Accept: application/json'"
)]
header: Option<Vec<String>>,
#[arg(
long = "filtered-query-param",
required = false,
help = "Filter the query parameters of the downloaded URL. If the download URL is the same, it will be scheduled as the same task. Examples: --filtered-query-param='signature' --filtered-query-param='timeout'"
help = "Filter the query parameters of the downloaded URL. If the download URL is the same, it will be scheduled as the same task, e.g. --filtered-query-param='signature' --filtered-query-param='timeout'"
)]
filtered_query_params: Option<Vec<String>>,
#[arg(
long = "include-files",
required = false,
help = "Filter files to download in a directory using glob patterns relative to the root URL's path. Examples: --include-files='*.txt' --include-files='subdir/file.txt'"
)]
include_files: Option<Vec<String>>,
#[arg(
long = "disable-back-to-source",
default_value_t = false,
@ -309,9 +305,6 @@ async fn main() -> anyhow::Result<()> {
args.log_max_files,
None,
None,
None,
None,
None,
false,
args.console,
);
@ -603,12 +596,7 @@ async fn main() -> anyhow::Result<()> {
Ok(())
}
/// Runs the dfget command to download files or directories from a given URL.
///
/// This function serves as the main entry point for the dfget download operation.
/// It handles both single file downloads and directory downloads based on the URL format.
/// The function performs path normalization, validates the URL scheme's capabilities,
/// and delegates to the appropriate download handler.
/// run runs the dfget command.
async fn run(mut args: Args, dfdaemon_download_client: DfdaemonDownloadClient) -> Result<()> {
// Get the absolute path of the output file.
args.output = Path::new(&args.output).absolutize()?.into();
@ -618,7 +606,7 @@ async fn run(mut args: Args, dfdaemon_download_client: DfdaemonDownloadClient) -
// then download all files in the directory. Otherwise, download the single file.
let scheme = args.url.scheme();
if args.url.path().ends_with('/') {
if BackendFactory::unsupported_download_directory(scheme) {
if !BackendFactory::supported_download_directory(scheme) {
return Err(Error::Unsupported(format!("{} download directory", scheme)));
};
@ -628,13 +616,7 @@ async fn run(mut args: Args, dfdaemon_download_client: DfdaemonDownloadClient) -
download(args, ProgressBar::new(0), dfdaemon_download_client).await
}
/// Downloads all files in a directory from various storage backends (object storage, HDFS, etc.).
///
/// This function handles directory-based downloads by recursively fetching all entries
/// in the specified directory. It supports filtering files based on include patterns,
/// enforces download limits, and performs concurrent downloads with configurable
/// concurrency control. The function creates the necessary directory structure
/// locally and downloads files while preserving the remote directory hierarchy.
/// download_dir downloads all files in the directory.
async fn download_dir(args: Args, download_client: DfdaemonDownloadClient) -> Result<()> {
// Initialize the object storage config and the hdfs config.
let object_storage = Some(ObjectStorage {
@ -651,17 +633,12 @@ async fn download_dir(args: Args, download_client: DfdaemonDownloadClient) -> Re
delegation_token: args.hdfs_delegation_token.clone(),
});
// Get all entries in the directory.
let mut entries = get_entries(&args, object_storage, hdfs, download_client.clone()).await?;
if let Some(ref include_files) = args.include_files {
entries = filter_entries(&args.url, entries, include_files)?;
}
// If the entries is empty, then return directly.
// Get all entries in the directory. If the directory is empty, then return directly.
let entries = get_entries(args.clone(), object_storage, hdfs).await?;
if entries.is_empty() {
warn!("no entries found in directory {}", args.url);
warn!("directory {} is empty", args.url);
return Ok(());
}
};
// If the actual file count is greater than the max_files, then reject the downloading.
let count = entries.iter().filter(|entry| !entry.is_dir).count();
@ -732,13 +709,7 @@ async fn download_dir(args: Args, download_client: DfdaemonDownloadClient) -> Re
Ok(())
}
/// Downloads a single file from various storage backends using the dfdaemon service.
///
/// This function handles single file downloads by communicating with a dfdaemon client.
/// It supports multiple storage protocols (object storage, HDFS, HTTP/HTTPS) and provides
/// two transfer modes: direct download by dfdaemon or streaming piece content through
/// the client. The function includes progress tracking, file creation, and proper error
/// handling throughout the download process.
/// download downloads the single file.
async fn download(
args: Args,
progress_bar: ProgressBar,
@ -785,7 +756,7 @@ async fn download(
.download_task(DownloadTaskRequest {
download: Some(Download {
url: args.url.to_string(),
digest: args.digest,
digest: Some(args.digest),
// NOTE: Dfget does not support range download.
range: None,
r#type: TaskType::Standard as i32,
@ -808,9 +779,9 @@ async fn download(
need_piece_content,
object_storage,
hdfs,
load_to_cache: false,
force_hard_link: args.force_hard_link,
content_for_calculating_task_id: args.content_for_calculating_task_id,
remote_ip: Some(local_ip().unwrap().to_string()),
}),
})
.await
@ -910,116 +881,69 @@ async fn download(
Ok(())
}
/// Retrieves all directory entries from a remote storage location.
///
/// This function communicates with the dfdaemon service to list all entries
/// (files and subdirectories) in the specified directory URL. It supports
/// various storage backends including object storage and HDFS by passing
/// the appropriate credentials and configuration. The function converts
/// the gRPC response into a local `DirEntry` format for further processing.
/// get_entries gets all entries in the directory.
async fn get_entries(
args: &Args,
args: Args,
object_storage: Option<ObjectStorage>,
hdfs: Option<Hdfs>,
download_client: DfdaemonDownloadClient,
) -> Result<Vec<DirEntry>> {
info!("list task entries: {:?}", args.url);
// List task entries.
let response = download_client
.list_task_entries(ListTaskEntriesRequest {
// Initialize backend factory and build backend.
let backend_factory = BackendFactory::new(None)?;
let backend = backend_factory.build(args.url.as_str())?;
// Collect backend request started metrics.
collect_backend_request_started_metrics(backend.scheme().as_str(), http::Method::HEAD.as_str());
// Record the start time.
let start_time = Instant::now();
let response = backend
.head(HeadRequest {
// NOTE: Mock a task id for head request.
task_id: Uuid::new_v4().to_string(),
url: args.url.to_string(),
request_header: header_vec_to_hashmap(args.header.clone().unwrap_or_default())?,
timeout: None,
certificate_chain: Vec::new(),
http_header: Some(header_vec_to_headermap(
args.header.clone().unwrap_or_default(),
)?),
timeout: args.timeout,
client_cert: None,
object_storage,
hdfs,
remote_ip: Some(local_ip().unwrap().to_string()),
})
.await
.inspect_err(|err| {
error!("list task entries failed: {}", err);
.inspect_err(|_err| {
// Collect backend request failure metrics.
collect_backend_request_failure_metrics(
backend.scheme().as_str(),
http::Method::HEAD.as_str(),
);
})?;
Ok(response
.entries
.into_iter()
.map(|entry| DirEntry {
url: entry.url,
content_length: entry.content_length as usize,
is_dir: entry.is_dir,
})
.collect())
}
// Return error when response is failed.
if !response.success {
// Collect backend request failure metrics.
collect_backend_request_failure_metrics(
backend.scheme().as_str(),
http::Method::HEAD.as_str(),
);
/// Filters directory entries based on include patterns and validates their URLs.
///
/// This function takes a collection of directory entries and filters them based on
/// glob patterns specified in `include_files`. It performs URL validation to ensure
/// all entries have valid URLs and that their paths fall within the scope of the
/// root URL. When an entry matches a pattern, both the entry and its parent
/// directory (if it exists) are included in the result.
fn filter_entries(
url: &Url,
entries: Vec<DirEntry>,
include_files: &[String],
) -> Result<Vec<DirEntry>> {
let patterns: Vec<Pattern> = include_files
.iter()
.filter_map(|include_file| Pattern::new(include_file).ok())
.collect();
// Build a HashMap of DirEntry objects keyed by relative paths for filtering and
// validates URLs and ensures paths are within the root URL's scope.
let mut entries_by_relative_path = HashMap::with_capacity(entries.len());
for entry in entries {
let entry_url = Url::parse(&entry.url).map_err(|err| {
error!("failed to parse entry URL '{}': {}", entry.url, err);
Error::ValidationError(format!("invalid URL: {}", entry.url))
})?;
let entry_path = entry_url.path();
match entry_path.strip_prefix(url.path()) {
Some(relative_path) => entries_by_relative_path
.insert(relative_path.trim_start_matches('/').to_string(), entry),
None => {
error!(
"entry path '{}' does not belong to the root path",
entry_path
);
return Err(Error::ValidationError(format!(
"path '{}' is outside the expected scope",
entry_path
)));
}
};
return Err(Error::BackendError(Box::new(BackendError {
message: response.error_message.unwrap_or_default(),
status_code: Some(response.http_status_code.unwrap_or_default()),
header: Some(response.http_header.unwrap_or_default()),
})));
}
// Filter entries by matching relative paths against patterns, including
// parent directories for matches.
let mut filtered_entries = HashSet::new();
for (relative_path, entry) in &entries_by_relative_path {
if patterns.iter().any(|pat| pat.matches(relative_path)) {
filtered_entries.insert(entry.clone());
if let Some(parent) = std::path::Path::new(relative_path).parent() {
if let Some(parent_entry) =
entries_by_relative_path.get(&parent.join("").to_string_lossy().to_string())
{
filtered_entries.insert(parent_entry.clone());
}
}
}
}
// Collect backend request finished metrics.
collect_backend_request_finished_metrics(
backend.scheme().as_str(),
http::Method::HEAD.as_str(),
start_time.elapsed(),
);
Ok(filtered_entries.into_iter().collect())
Ok(response.entries)
}
/// Constructs the local output path for a directory entry based on its remote URL.
///
/// This function maps a remote directory entry to its corresponding local file system
/// path by replacing the remote root directory with the local output directory.
/// It handles URL percent-decoding to ensure proper path construction and maintains
/// the relative directory structure from the remote source.
/// make_output_by_entry makes the output path by the entry information.
fn make_output_by_entry(url: Url, output: &Path, entry: DirEntry) -> Result<PathBuf> {
// Get the root directory of the download directory and the output root directory.
let root_dir = url.path().to_string();
@ -1037,12 +961,7 @@ fn make_output_by_entry(url: Url, output: &Path, entry: DirEntry) -> Result<Path
.into())
}
/// Creates and validates a dfdaemon download client with health checking.
///
/// This function establishes a connection to the dfdaemon service via Unix domain socket
/// and performs a health check to ensure the service is running and ready to handle
/// download requests. Only after successful health verification does it return the
/// download client for actual use.
/// get_and_check_dfdaemon_download_client gets a dfdaemon download client and checks its health.
async fn get_dfdaemon_download_client(endpoint: PathBuf) -> Result<DfdaemonDownloadClient> {
// Check dfdaemon's health.
let health_client = HealthClient::new_unix(endpoint.clone()).await?;
@ -1053,13 +972,7 @@ async fn get_dfdaemon_download_client(endpoint: PathBuf) -> Result<DfdaemonDownl
Ok(dfdaemon_download_client)
}
/// Validates command line arguments for consistency and safety requirements.
///
/// This function performs comprehensive validation of the download arguments to ensure
/// they are logically consistent and safe to execute. It checks URL-output path matching,
/// directory existence, file conflicts, piece length constraints, and glob pattern validity.
/// The validation prevents common user errors and potential security issues before
/// starting the download process.
/// validate_args validates the command line arguments.
fn validate_args(args: &Args) -> Result<()> {
// If the URL is a directory, the output path should be a directory.
if args.url.path().ends_with('/') && !args.output.is_dir() {
@ -1108,42 +1021,9 @@ fn validate_args(args: &Args) -> Result<()> {
}
}
if let Some(ref include_files) = args.include_files {
for include_file in include_files {
if Pattern::new(include_file).is_err() {
return Err(Error::ValidationError(format!(
"invalid glob pattern in include_files: '{}'",
include_file
)));
}
if !is_normal_relative_path(include_file) {
return Err(Error::ValidationError(format!(
"path is not a normal relative path in include_files: '{}'. It must not contain '..', '.', or start with '/'.",
include_file
)));
}
}
}
Ok(())
}
/// Validates that a path string is a normal relative path without unsafe components.
///
/// This function ensures that a given path is both relative (doesn't start with '/')
/// and contains only normal path components. It rejects paths with parent directory
/// references ('..'), current directory references ('.'), or any other special
/// path components that could be used for directory traversal attacks or unexpected
/// file system navigation.
fn is_normal_relative_path(path: &str) -> bool {
let path = Path::new(path);
path.is_relative()
&& path
.components()
.all(|comp| matches!(comp, Component::Normal(_)))
}
#[cfg(test)]
mod tests {
use super::*;
@ -1307,346 +1187,4 @@ mod tests {
let result = make_output_by_entry(url, output, entry);
assert!(result.is_err());
}
#[test]
fn should_filter_entries() {
let test_cases = vec![
(
Url::parse("http://example.com/root/").unwrap(),
vec![
DirEntry {
url: "http://example.com/root/dir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/file.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/file2.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file3.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file4.png".to_string(),
content_length: 100,
is_dir: false,
},
],
vec!["dir/file.txt".to_string()],
vec![
DirEntry {
url: "http://example.com/root/dir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/file.txt".to_string(),
content_length: 100,
is_dir: false,
},
],
),
(
Url::parse("http://example.com/root/").unwrap(),
vec![
DirEntry {
url: "http://example.com/root/dir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/file.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/file2.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file3.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file4.png".to_string(),
content_length: 100,
is_dir: false,
},
],
vec![
"dir/file.txt".to_string(),
"dir/subdir/file4.png".to_string(),
],
vec![
DirEntry {
url: "http://example.com/root/dir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/file.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file4.png".to_string(),
content_length: 100,
is_dir: false,
},
],
),
(
Url::parse("http://example.com/root/").unwrap(),
vec![
DirEntry {
url: "http://example.com/root/dir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/file.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/file2.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file3.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file4.png".to_string(),
content_length: 100,
is_dir: false,
},
],
vec!["dir/subdir/*.png".to_string()],
vec![
DirEntry {
url: "http://example.com/root/dir/subdir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file4.png".to_string(),
content_length: 100,
is_dir: false,
},
],
),
(
Url::parse("http://example.com/root/").unwrap(),
vec![
DirEntry {
url: "http://example.com/root/dir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/file.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/file2.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file3.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file4.png".to_string(),
content_length: 100,
is_dir: false,
},
],
vec!["dir/*".to_string()],
vec![
DirEntry {
url: "http://example.com/root/dir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/file.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/file2.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file3.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file4.png".to_string(),
content_length: 100,
is_dir: false,
},
],
),
(
Url::parse("http://example.com/root/").unwrap(),
vec![
DirEntry {
url: "http://example.com/root/dir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/file.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/file2.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file3.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file4.png".to_string(),
content_length: 100,
is_dir: false,
},
],
vec!["dir/".to_string()],
vec![DirEntry {
url: "http://example.com/root/dir/".to_string(),
content_length: 10,
is_dir: true,
}],
),
(
Url::parse("http://example.com/root/").unwrap(),
vec![
DirEntry {
url: "http://example.com/root/dir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/file.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/file2.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/".to_string(),
content_length: 10,
is_dir: true,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file3.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: "http://example.com/root/dir/subdir/file4.png".to_string(),
content_length: 100,
is_dir: false,
},
],
vec!["test".to_string()],
vec![],
),
(
Url::parse("http://example.com/root/").unwrap(),
vec![
DirEntry {
url: "http://example.com/root/dir/file.txt".to_string(),
content_length: 100,
is_dir: false,
},
DirEntry {
url: " ".to_string(),
content_length: 100,
is_dir: false,
},
],
vec!["dir/file.txt".to_string()],
vec![],
),
];
for (url, entries, include_files, expected_entries) in test_cases {
let result = filter_entries(&url, entries, &include_files);
if result.is_err() {
assert!(matches!(result, Err(Error::ValidationError(_))));
} else {
let filtered_entries = result.unwrap();
assert_eq!(filtered_entries.len(), expected_entries.len());
for filtered_entry in &filtered_entries {
assert!(expected_entries
.iter()
.any(|expected_entry| { expected_entry.url == filtered_entry.url }));
}
}
}
}
}

View File

@ -162,7 +162,6 @@ impl Dynconfig {
location: self.config.host.location.clone(),
version: CARGO_PKG_VERSION.to_string(),
commit: GIT_COMMIT_SHORT_HASH.to_string(),
scheduler_cluster_id: self.config.host.scheduler_cluster_id.unwrap_or(0),
})
.await
}

View File

@ -18,35 +18,30 @@ use crate::metrics::{
collect_delete_host_failure_metrics, collect_delete_host_started_metrics,
collect_delete_task_failure_metrics, collect_delete_task_started_metrics,
collect_download_task_failure_metrics, collect_download_task_finished_metrics,
collect_download_task_started_metrics, collect_list_task_entries_failure_metrics,
collect_list_task_entries_started_metrics, collect_stat_task_failure_metrics,
collect_download_task_started_metrics, collect_stat_task_failure_metrics,
collect_stat_task_started_metrics, collect_upload_task_failure_metrics,
collect_upload_task_finished_metrics, collect_upload_task_started_metrics,
};
use crate::resource::{persistent_cache_task, task};
use crate::shutdown;
use dragonfly_api::common::v2::{CacheTask, PersistentCacheTask, Priority, Task, TaskType};
use dragonfly_api::common::v2::{PersistentCacheTask, Priority, Task, TaskType};
use dragonfly_api::dfdaemon::v2::{
dfdaemon_download_client::DfdaemonDownloadClient as DfdaemonDownloadGRPCClient,
dfdaemon_download_server::{
DfdaemonDownload, DfdaemonDownloadServer as DfdaemonDownloadGRPCServer,
},
DeleteCacheTaskRequest, DeleteTaskRequest, DownloadCacheTaskRequest, DownloadCacheTaskResponse,
DownloadPersistentCacheTaskRequest, DownloadPersistentCacheTaskResponse, DownloadTaskRequest,
DownloadTaskResponse, Entry, ListTaskEntriesRequest, ListTaskEntriesResponse,
StatCacheTaskRequest as DfdaemonStatCacheTaskRequest, StatPersistentCacheTaskRequest,
DeleteTaskRequest, DownloadPersistentCacheTaskRequest, DownloadPersistentCacheTaskResponse,
DownloadTaskRequest, DownloadTaskResponse, StatPersistentCacheTaskRequest,
StatTaskRequest as DfdaemonStatTaskRequest, UploadPersistentCacheTaskRequest,
};
use dragonfly_api::errordetails::v2::Backend;
use dragonfly_api::scheduler::v2::DeleteHostRequest as SchedulerDeleteHostRequest;
use dragonfly_client_backend::HeadRequest;
use dragonfly_client_config::dfdaemon::Config;
use dragonfly_client_core::{
error::{ErrorType, OrErr},
Error as ClientError, Result as ClientResult,
};
use dragonfly_client_util::{
digest::{verify_file_digest, Digest},
http::{get_range, hashmap_to_headermap, headermap_to_hashmap},
id_generator::{PersistentCacheTaskIDParameter, TaskIDParameter},
};
@ -120,7 +115,6 @@ impl DfdaemonDownloadServer {
// Initialize the grpc service.
let service = DfdaemonDownloadGRPCServer::with_interceptor(
DfdaemonDownloadServerHandler {
config: self.config.clone(),
socket_path: self.socket_path.clone(),
task: self.task.clone(),
persistent_cache_task: self.persistent_cache_task.clone(),
@ -139,6 +133,11 @@ impl DfdaemonDownloadServer {
// Initialize health reporter.
let (mut health_reporter, health_service) = tonic_health::server::health_reporter();
// Set the serving status of the download grpc server.
health_reporter
.set_serving::<DfdaemonDownloadGRPCServer<DfdaemonDownloadServerHandler>>()
.await;
// Start download grpc server with unix domain socket.
fs::create_dir_all(self.socket_path.parent().unwrap()).await?;
fs::remove_file(self.socket_path.clone())
@ -149,7 +148,7 @@ impl DfdaemonDownloadServer {
// Bind the unix domain socket and set the permissions for the socket.
let uds = UnixListener::bind(&self.socket_path)?;
let perms = std::fs::Permissions::from_mode(0o777);
let perms = std::fs::Permissions::from_mode(0o660);
fs::set_permissions(&self.socket_path, perms).await?;
// TODO(Gaius): RateLimitLayer is not implemented Clone, so we can't use it here.
@ -166,7 +165,7 @@ impl DfdaemonDownloadServer {
.http2_keepalive_interval(Some(super::HTTP2_KEEP_ALIVE_INTERVAL))
.http2_keepalive_timeout(Some(super::HTTP2_KEEP_ALIVE_TIMEOUT))
.layer(rate_limit_layer)
.add_service(reflection)
.add_service(reflection.clone())
.add_service(health_service)
.add_service(service)
.serve_with_incoming_shutdown(uds_stream, async move {
@ -176,12 +175,6 @@ impl DfdaemonDownloadServer {
// Notify the download grpc server is started.
_ = grpc_server_started_barrier.wait() => {
info!("download server is ready to start");
health_reporter
.set_serving::<DfdaemonDownloadGRPCServer<DfdaemonDownloadServerHandler>>()
.await;
info!("download server's health status set to serving");
}
// Wait for shutdown signal.
_ = shutdown.recv() => {
@ -211,9 +204,6 @@ impl DfdaemonDownloadServer {
/// DfdaemonDownloadServerHandler is the handler of the dfdaemon download grpc service.
pub struct DfdaemonDownloadServerHandler {
/// config is the configuration of the dfdaemon.
config: Arc<Config>,
/// socket_path is the path of the unix domain socket.
socket_path: PathBuf,
@ -231,10 +221,7 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
type DownloadTaskStream = ReceiverStream<Result<DownloadTaskResponse, Status>>;
/// download_task tells the dfdaemon to download the task.
#[instrument(
skip_all,
fields(host_id, task_id, peer_id, url, remote_ip, content_length)
)]
#[instrument(skip_all, fields(host_id, task_id, peer_id))]
async fn download_task(
&self,
request: Request<DownloadTaskRequest>,
@ -285,11 +272,6 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
Span::current().record("host_id", host_id.as_str());
Span::current().record("task_id", task_id.as_str());
Span::current().record("peer_id", peer_id.as_str());
Span::current().record("url", download.url.clone());
Span::current().record(
"remote_ip",
download.remote_ip.clone().unwrap_or_default().as_str(),
);
info!("download task in download server");
// Download task started.
@ -363,15 +345,12 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
error!("missing content length in the response");
return Err(Status::internal("missing content length in the response"));
};
info!(
"content length {}, piece length {}",
content_length,
task.piece_length().unwrap_or_default()
);
Span::current().record("content_length", content_length);
// Download's range priority is higher than the request header's range.
// If download protocol is http, use the range of the request header.
// If download protocol is not http, use the range of the download.
@ -504,48 +483,22 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
)),
)
.await;
return;
}
Err(err) => {
error!("check output path: {}", err);
handle_error(&out_stream_tx, err).await;
return;
}
}
} else if let Err(err) = task_manager_clone
return;
}
if let Err(err) = task_manager_clone
.copy_task(task_clone.id.as_str(), output_path)
.await
{
error!("copy task: {}", err);
handle_error(&out_stream_tx, err).await;
return;
}
}
// Verify the file digest if it is provided.
if let Some(raw_digest) = &download_clone.digest {
let digest = match raw_digest.parse::<Digest>() {
Ok(digest) => digest,
Err(err) => {
error!("parse digest: {}", err);
handle_error(
&out_stream_tx,
Status::invalid_argument(format!(
"invalid digest({}): {}",
raw_digest, err
)),
)
.await;
return;
}
};
if let Err(err) =
verify_file_digest(digest, Path::new(output_path.as_str()))
{
error!("verify file digest: {}", err);
handle_error(&out_stream_tx, err).await;
return;
}
}
}
@ -670,7 +623,7 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
}
/// stat_task gets the status of the task.
#[instrument(skip_all, fields(host_id, task_id, remote_ip, local_only))]
#[instrument(skip_all, fields(host_id, task_id))]
async fn stat_task(
&self,
request: Request<DfdaemonStatTaskRequest>,
@ -689,128 +642,32 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
// Get the task id from the request.
let task_id = request.task_id;
// Get the local_only flag from the request, default to false.
let local_only = request.local_only;
// Span record the host id and task id.
Span::current().record("host_id", host_id.as_str());
Span::current().record("task_id", task_id.as_str());
Span::current().record(
"remote_ip",
request.remote_ip.clone().unwrap_or_default().as_str(),
);
Span::current().record("local_only", local_only.to_string().as_str());
info!("stat task in download server");
// Collect the stat task metrics.
collect_stat_task_started_metrics(TaskType::Standard as i32);
match self
// Get the task from the scheduler.
let task = self
.task
.stat(task_id.as_str(), host_id.as_str(), local_only)
.stat(task_id.as_str(), host_id.as_str())
.await
{
Ok(task) => Ok(Response::new(task)),
Err(err) => {
.map_err(|err| {
// Collect the stat task failure metrics.
collect_stat_task_failure_metrics(TaskType::Standard as i32);
// Log the error with detailed context.
error!("stat task failed: {}", err);
// Map the error to an appropriate gRPC status.
Err(match err {
ClientError::TaskNotFound(id) => {
Status::not_found(format!("task not found: {}", id))
}
_ => Status::internal(err.to_string()),
})
}
}
}
/// list_tasks lists the tasks.
#[instrument(skip_all, fields(task_id, url, remote_ip))]
async fn list_task_entries(
&self,
request: Request<ListTaskEntriesRequest>,
) -> Result<Response<ListTaskEntriesResponse>, Status> {
// If the parent context is set, use it as the parent context for the span.
if let Some(parent_ctx) = request.extensions().get::<Context>() {
Span::current().set_parent(parent_ctx.clone());
};
// Clone the request.
let request = request.into_inner();
// Span record the task id and url.
Span::current().record("task_id", request.task_id.as_str());
Span::current().record("url", request.url.as_str());
Span::current().record(
"remote_ip",
request.remote_ip.clone().unwrap_or_default().as_str(),
);
info!("list tasks in download server");
// Collect the list tasks started metrics.
collect_list_task_entries_started_metrics(TaskType::Standard as i32);
// Build the backend.
let backend = self
.task
.backend_factory
.build(request.url.as_str())
.map_err(|err| {
// Collect the list tasks failure metrics.
collect_list_task_entries_failure_metrics(TaskType::Standard as i32);
error!("build backend: {}", err);
error!("stat task: {}", err);
Status::internal(err.to_string())
})?;
// Head the task entries.
let response = backend
.head(HeadRequest {
task_id: request.task_id.clone(),
url: request.url.clone(),
http_header: Some(hashmap_to_headermap(&request.request_header).map_err(
|err| {
error!("parse request header: {}", err);
Status::internal(err.to_string())
},
)?),
timeout: self.config.download.piece_timeout,
client_cert: None,
object_storage: request.object_storage.clone(),
hdfs: request.hdfs.clone(),
})
.await
.map_err(|err| {
// Collect the list tasks failure metrics.
collect_list_task_entries_failure_metrics(TaskType::Standard as i32);
error!("list task entries: {}", err);
Status::internal(err.to_string())
})?;
Ok(Response::new(ListTaskEntriesResponse {
content_length: response.content_length.unwrap_or_default(),
response_header: headermap_to_hashmap(&response.http_header.unwrap_or_default()),
status_code: response.http_status_code.map(|code| code.as_u16().into()),
entries: response
.entries
.into_iter()
.map(|dir_entry| Entry {
url: dir_entry.url,
content_length: dir_entry.content_length as u64,
is_dir: dir_entry.is_dir,
})
.collect(),
}))
Ok(Response::new(task))
}
/// delete_task calls the dfdaemon to delete the task.
#[instrument(skip_all, fields(host_id, task_id, remote_ip))]
#[instrument(skip_all, fields(host_id, task_id))]
async fn delete_task(
&self,
request: Request<DeleteTaskRequest>,
@ -832,10 +689,6 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
// Span record the host id and task id.
Span::current().record("host_id", host_id.as_str());
Span::current().record("task_id", task_id.as_str());
Span::current().record(
"remote_ip",
request.remote_ip.clone().unwrap_or_default().as_str(),
);
info!("delete task in download server");
// Collect the delete task started metrics.
@ -894,7 +747,7 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
ReceiverStream<Result<DownloadPersistentCacheTaskResponse, Status>>;
/// download_persistent_cache_task downloads the persistent cache task.
#[instrument(skip_all, fields(host_id, task_id, peer_id, remote_ip, content_length))]
#[instrument(skip_all, fields(host_id, task_id, peer_id))]
async fn download_persistent_cache_task(
&self,
request: Request<DownloadPersistentCacheTaskRequest>,
@ -927,10 +780,6 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
Span::current().record("host_id", host_id.as_str());
Span::current().record("task_id", task_id.as_str());
Span::current().record("peer_id", peer_id.as_str());
Span::current().record(
"remote_ip",
request.remote_ip.clone().unwrap_or_default().as_str(),
);
info!("download persistent cache task in download server");
// Download task started.
@ -984,15 +833,12 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
task
}
};
info!(
"content length {}, piece length {}",
task.content_length(),
task.piece_length()
);
Span::current().record("content_length", task.content_length());
// Initialize stream channel.
let request_clone = request.clone();
let task_manager_clone = self.persistent_cache_task.clone();
@ -1074,48 +920,22 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
)),
)
.await;
return;
}
Err(err) => {
error!("check output path: {}", err);
handle_error(&out_stream_tx, err).await;
return;
}
}
} else if let Err(err) = task_manager_clone
return;
}
if let Err(err) = task_manager_clone
.copy_task(task_clone.id.as_str(), output_path)
.await
{
error!("copy task: {}", err);
handle_error(&out_stream_tx, err).await;
return;
}
}
// Verify the file digest if it is provided.
if let Some(raw_digest) = &request_clone.digest {
let digest = match raw_digest.parse::<Digest>() {
Ok(digest) => digest,
Err(err) => {
error!("parse digest: {}", err);
handle_error(
&out_stream_tx,
Status::invalid_argument(format!(
"invalid digest({}): {}",
raw_digest, err
)),
)
.await;
return;
}
};
if let Err(err) =
verify_file_digest(digest, Path::new(output_path.as_str()))
{
error!("verify file digest: {}", err);
handle_error(&out_stream_tx, err).await;
return;
}
}
}
@ -1152,7 +972,7 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
}
/// upload_persistent_cache_task uploads the persistent cache task.
#[instrument(skip_all, fields(host_id, task_id, peer_id, remote_ip))]
#[instrument(skip_all, fields(host_id, task_id, peer_id))]
async fn upload_persistent_cache_task(
&self,
request: Request<UploadPersistentCacheTaskRequest>,
@ -1199,10 +1019,6 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
Span::current().record("host_id", host_id.as_str());
Span::current().record("task_id", task_id.as_str());
Span::current().record("peer_id", peer_id.as_str());
Span::current().record(
"remote_ip",
request.remote_ip.clone().unwrap_or_default().as_str(),
);
info!("upload persistent cache task in download server");
// Collect upload task started metrics.
@ -1253,7 +1069,7 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
}
/// stat_persistent_cache_task stats the persistent cache task.
#[instrument(skip_all, fields(host_id, task_id, remote_ip))]
#[instrument(skip_all, fields(host_id, task_id))]
async fn stat_persistent_cache_task(
&self,
request: Request<StatPersistentCacheTaskRequest>,
@ -1275,10 +1091,6 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
// Span record the host id and task id.
Span::current().record("host_id", host_id.as_str());
Span::current().record("task_id", task_id.as_str());
Span::current().record(
"remote_ip",
request.remote_ip.clone().unwrap_or_default().as_str(),
);
info!("stat persistent cache task in download server");
// Collect the stat persistent cache task started metrics.
@ -1298,39 +1110,6 @@ impl DfdaemonDownload for DfdaemonDownloadServerHandler {
Ok(Response::new(task))
}
/// DownloadCacheTaskStream is the stream of the download cache task response.
type DownloadCacheTaskStream = ReceiverStream<Result<DownloadCacheTaskResponse, Status>>;
/// download_cache_task tells the dfdaemon to download the cache task.
#[instrument(
skip_all,
fields(host_id, task_id, peer_id, url, remote_ip, content_length)
)]
async fn download_cache_task(
&self,
_request: Request<DownloadCacheTaskRequest>,
) -> Result<Response<Self::DownloadCacheTaskStream>, Status> {
todo!();
}
/// stat_cache_task gets the status of the cache task.
#[instrument(skip_all, fields(host_id, task_id, remote_pi, local_only))]
async fn stat_cache_task(
&self,
_request: Request<DfdaemonStatCacheTaskRequest>,
) -> Result<Response<CacheTask>, Status> {
todo!();
}
/// delete_cache_task calls the dfdaemon to delete the cache task.
#[instrument(skip_all, fields(host_id, task_id, remote_ip))]
async fn delete_cache_task(
&self,
_request: Request<DeleteCacheTaskRequest>,
) -> Result<Response<()>, Status> {
todo!();
}
}
/// DfdaemonDownloadClient is a wrapper of DfdaemonDownloadGRPCClient.
@ -1408,18 +1187,6 @@ impl DfdaemonDownloadClient {
Ok(response.into_inner())
}
/// list_task_entries lists the task entries.
#[instrument(skip_all)]
pub async fn list_task_entries(
&self,
request: ListTaskEntriesRequest,
) -> ClientResult<ListTaskEntriesResponse> {
let request = Self::make_request(request);
info!("list task entries request: {:?}", request);
let response = self.client.clone().list_task_entries(request).await?;
Ok(response.into_inner())
}
/// delete_task tells the dfdaemon to delete the task.
#[instrument(skip_all)]
pub async fn delete_task(&self, request: DeleteTaskRequest) -> ClientResult<()> {

View File

@ -24,22 +24,20 @@ use crate::metrics::{
};
use crate::resource::{persistent_cache_task, task};
use crate::shutdown;
use bytesize::MB;
use dragonfly_api::common::v2::{
CacheTask, Host, Network, PersistentCacheTask, Piece, Priority, Task, TaskType,
Host, Network, PersistentCacheTask, Piece, Priority, Task, TaskType,
};
use dragonfly_api::dfdaemon::v2::{
dfdaemon_upload_client::DfdaemonUploadClient as DfdaemonUploadGRPCClient,
dfdaemon_upload_server::{DfdaemonUpload, DfdaemonUploadServer as DfdaemonUploadGRPCServer},
DeleteCacheTaskRequest, DeletePersistentCacheTaskRequest, DeleteTaskRequest,
DownloadCachePieceRequest, DownloadCachePieceResponse, DownloadCacheTaskRequest,
DownloadCacheTaskResponse, DownloadPersistentCachePieceRequest,
DeletePersistentCacheTaskRequest, DeleteTaskRequest, DownloadPersistentCachePieceRequest,
DownloadPersistentCachePieceResponse, DownloadPersistentCacheTaskRequest,
DownloadPersistentCacheTaskResponse, DownloadPieceRequest, DownloadPieceResponse,
DownloadTaskRequest, DownloadTaskResponse, ExchangeIbVerbsQueuePairEndpointRequest,
ExchangeIbVerbsQueuePairEndpointResponse, StatCacheTaskRequest, StatPersistentCacheTaskRequest,
StatTaskRequest, SyncCachePiecesRequest, SyncCachePiecesResponse, SyncHostRequest,
SyncPersistentCachePiecesRequest, SyncPersistentCachePiecesResponse, SyncPiecesRequest,
SyncPiecesResponse, UpdatePersistentCacheTaskRequest,
ExchangeIbVerbsQueuePairEndpointResponse, StatPersistentCacheTaskRequest, StatTaskRequest,
SyncHostRequest, SyncPersistentCachePiecesRequest, SyncPersistentCachePiecesResponse,
SyncPiecesRequest, SyncPiecesResponse, UpdatePersistentCacheTaskRequest,
};
use dragonfly_api::errordetails::v2::Backend;
use dragonfly_client_config::dfdaemon::Config;
@ -50,13 +48,14 @@ use dragonfly_client_core::{
use dragonfly_client_util::{
http::{get_range, hashmap_to_headermap, headermap_to_hashmap},
id_generator::TaskIDParameter,
net::Interface,
net::{get_interface_info, Interface},
};
use opentelemetry::Context;
use std::net::SocketAddr;
use std::path::{Path, PathBuf};
use std::sync::Arc;
use std::time::{Duration, Instant};
use sysinfo::Networks;
use tokio::io::AsyncReadExt;
use tokio::sync::mpsc;
use tokio::sync::mpsc::Sender;
@ -88,9 +87,6 @@ pub struct DfdaemonUploadServer {
/// persistent_cache_task is the persistent cache task manager.
persistent_cache_task: Arc<persistent_cache_task::PersistentCacheTask>,
/// interface is the network interface.
interface: Arc<Interface>,
/// shutdown is used to shutdown the grpc server.
shutdown: shutdown::Shutdown,
@ -106,7 +102,6 @@ impl DfdaemonUploadServer {
addr: SocketAddr,
task: Arc<task::Task>,
persistent_cache_task: Arc<persistent_cache_task::PersistentCacheTask>,
interface: Arc<Interface>,
shutdown: shutdown::Shutdown,
shutdown_complete_tx: mpsc::UnboundedSender<()>,
) -> Self {
@ -114,7 +109,6 @@ impl DfdaemonUploadServer {
config,
addr,
task,
interface,
persistent_cache_task,
shutdown,
_shutdown_complete: shutdown_complete_tx,
@ -123,12 +117,17 @@ impl DfdaemonUploadServer {
/// run starts the upload server.
pub async fn run(&mut self, grpc_server_started_barrier: Arc<Barrier>) -> ClientResult<()> {
// Initialize the grpc service.
let interface =
get_interface_info(self.config.host.ip.unwrap(), self.config.upload.rate_limit)
.unwrap();
let service = DfdaemonUploadGRPCServer::with_interceptor(
DfdaemonUploadServerHandler {
interface,
socket_path: self.config.download.server.socket_path.clone(),
task: self.task.clone(),
persistent_cache_task: self.persistent_cache_task.clone(),
interface: self.interface.clone(),
},
ExtractTracingInterceptor,
);
@ -144,6 +143,11 @@ impl DfdaemonUploadServer {
// Initialize health reporter.
let (mut health_reporter, health_service) = tonic_health::server::health_reporter();
// Set the serving status of the upload grpc server.
health_reporter
.set_serving::<DfdaemonUploadGRPCServer<DfdaemonUploadServerHandler>>()
.await;
// TODO(Gaius): RateLimitLayer is not implemented Clone, so we can't use it here.
// Only use the LoadShed layer and the ConcurrencyLimit layer.
let rate_limit_layer = ServiceBuilder::new()
@ -167,7 +171,7 @@ impl DfdaemonUploadServer {
.http2_keepalive_interval(Some(super::HTTP2_KEEP_ALIVE_INTERVAL))
.http2_keepalive_timeout(Some(super::HTTP2_KEEP_ALIVE_TIMEOUT))
.layer(rate_limit_layer)
.add_service(reflection)
.add_service(reflection.clone())
.add_service(health_service)
.add_service(service)
.serve_with_shutdown(self.addr, async move {
@ -176,13 +180,7 @@ impl DfdaemonUploadServer {
tokio::select! {
// Notify the upload grpc server is started.
_ = grpc_server_started_barrier.wait() => {
info!("upload server is ready to start");
health_reporter
.set_serving::<DfdaemonUploadGRPCServer<DfdaemonUploadServerHandler>>()
.await;
info!("upload server's health status set to serving");
info!("upload server is ready");
}
// Wait for shutdown signal.
_ = shutdown.recv() => {
@ -204,6 +202,9 @@ impl DfdaemonUploadServer {
/// DfdaemonUploadServerHandler is the handler of the dfdaemon upload grpc service.
pub struct DfdaemonUploadServerHandler {
/// interface is the network interface.
interface: Interface,
/// socket_path is the path of the unix domain socket.
socket_path: PathBuf,
@ -212,9 +213,6 @@ pub struct DfdaemonUploadServerHandler {
/// persistent_cache_task is the persistent cache task manager.
persistent_cache_task: Arc<persistent_cache_task::PersistentCacheTask>,
/// interface is the network interface.
interface: Arc<Interface>,
}
/// DfdaemonUploadServerHandler implements the dfdaemon upload grpc service.
@ -224,10 +222,7 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
type DownloadTaskStream = ReceiverStream<Result<DownloadTaskResponse, Status>>;
/// download_task downloads the task.
#[instrument(
skip_all,
fields(host_id, task_id, peer_id, url, remote_ip, content_length)
)]
#[instrument(skip_all, fields(host_id, task_id, peer_id))]
async fn download_task(
&self,
request: Request<DownloadTaskRequest>,
@ -278,11 +273,6 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
Span::current().record("host_id", host_id.as_str());
Span::current().record("task_id", task_id.as_str());
Span::current().record("peer_id", peer_id.as_str());
Span::current().record("url", download.url.clone());
Span::current().record(
"remote_ip",
download.remote_ip.clone().unwrap_or_default().as_str(),
);
info!("download task in upload server");
// Download task started.
@ -363,8 +353,6 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
task.piece_length().unwrap_or_default()
);
Span::current().record("content_length", content_length);
// Download's range priority is higher than the request header's range.
// If download protocol is http, use the range of the request header.
// If download protocol is not http, use the range of the download.
@ -639,7 +627,7 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
}
/// stat_task stats the task.
#[instrument(skip_all, fields(host_id, task_id, remote_ip, local_only))]
#[instrument(skip_all, fields(host_id, task_id))]
async fn stat_task(&self, request: Request<StatTaskRequest>) -> Result<Response<Task>, Status> {
// If the parent context is set, use it as the parent context for the span.
if let Some(parent_ctx) = request.extensions().get::<Context>() {
@ -655,48 +643,32 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
// Get the task id from the request.
let task_id = request.task_id;
// Get the local_only flag from the request, default to false.
let local_only = request.local_only;
// Span record the host id and task id.
Span::current().record("host_id", host_id.as_str());
Span::current().record("task_id", task_id.as_str());
Span::current().record(
"remote_ip",
request.remote_ip.clone().unwrap_or_default().as_str(),
);
Span::current().record("local_only", local_only.to_string().as_str());
info!("stat task in upload server");
// Collect the stat task metrics.
collect_stat_task_started_metrics(TaskType::Standard as i32);
match self
// Get the task from the scheduler.
let task = self
.task
.stat(task_id.as_str(), host_id.as_str(), local_only)
.stat(task_id.as_str(), host_id.as_str())
.await
{
Ok(task) => Ok(Response::new(task)),
Err(err) => {
.map_err(|err| {
// Collect the stat task failure metrics.
collect_stat_task_failure_metrics(TaskType::Standard as i32);
// Log the error with detailed context.
error!("stat task failed: {}", err);
error!("stat task: {}", err);
Status::internal(err.to_string())
})?;
// Map the error to an appropriate gRPC status.
Err(match err {
ClientError::TaskNotFound(id) => {
Status::not_found(format!("task not found: {}", id))
}
_ => Status::internal(err.to_string()),
})
}
}
Ok(Response::new(task))
}
/// delete_task deletes the task.
#[instrument(skip_all, fields(host_id, task_id, remote_ip))]
#[instrument(skip_all, fields(host_id, task_id))]
async fn delete_task(
&self,
request: Request<DeleteTaskRequest>,
@ -718,10 +690,6 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
// Span record the host id and task id.
Span::current().record("host_id", host_id.as_str());
Span::current().record("task_id", task_id.as_str());
Span::current().record(
"remote_ip",
request.remote_ip.clone().unwrap_or_default().as_str(),
);
info!("delete task in upload server");
// Collect the delete task started metrics.
@ -745,8 +713,7 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
/// SyncPiecesStream is the stream of the sync pieces response.
type SyncPiecesStream = ReceiverStream<Result<SyncPiecesResponse, Status>>;
/// sync_pieces provides the piece metadata for parent. If the per-piece collection timeout is exceeded,
/// the stream will be closed.
/// sync_pieces provides the piece metadata for parent.
#[instrument(skip_all, fields(host_id, remote_host_id, task_id))]
async fn sync_pieces(
&self,
@ -786,6 +753,7 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
tokio::spawn(
async move {
loop {
let mut has_started_piece = false;
let mut finished_piece_numbers = Vec::new();
for interested_piece_number in interested_piece_numbers.iter() {
let piece = match task_manager.piece.get(
@ -850,6 +818,11 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
finished_piece_numbers.push(piece.number);
continue;
}
// Check whether the piece is started.
if piece.is_started() {
has_started_piece = true;
}
}
// Remove the finished piece numbers from the interested piece numbers.
@ -863,6 +836,13 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
return;
}
// If there is no started piece, return.
if !has_started_piece {
info!("there is no started piece");
drop(out_stream_tx);
return;
}
// Wait for the piece to be finished.
tokio::time::sleep(
dragonfly_client_storage::DEFAULT_WAIT_FOR_PIECE_FINISHED_INTERVAL,
@ -877,10 +857,7 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
}
/// download_piece provides the piece content for parent.
#[instrument(
skip_all,
fields(host_id, remote_host_id, task_id, piece_id, piece_length)
)]
#[instrument(skip_all, fields(host_id, remote_host_id, task_id, piece_id))]
async fn download_piece(
&self,
request: Request<DownloadPieceRequest>,
@ -928,8 +905,6 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
Status::not_found("piece metadata not found")
})?;
Span::current().record("piece_length", piece.length);
// Collect upload piece started metrics.
collect_upload_piece_started_metrics();
info!("start upload piece content");
@ -963,7 +938,6 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
error!("upload piece content failed: {}", err);
Status::internal(err.to_string())
})?;
drop(reader);
// Collect upload piece finished metrics.
collect_upload_piece_finished_metrics();
@ -1002,6 +976,9 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
Span::current().set_parent(parent_ctx.clone());
};
// DEFAULT_HOST_INFO_REFRESH_INTERVAL is the default interval for refreshing the host info.
const DEFAULT_HOST_INFO_REFRESH_INTERVAL: Duration = Duration::from_millis(500);
// Clone the request.
let request = request.into_inner();
@ -1023,42 +1000,43 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
// Get local interface.
let interface = self.interface.clone();
// DEFAULT_HOST_INFO_REFRESH_INTERVAL is the default interval for refreshing the host info.
const DEFAULT_HOST_INFO_REFRESH_INTERVAL: Duration = Duration::from_millis(500);
// Initialize stream channel.
let (out_stream_tx, out_stream_rx) = mpsc::channel(10 * 1024);
tokio::spawn(
async move {
// Initialize sysinfo network.
let mut networks = Networks::new_with_refreshed_list();
// Start the host info update loop.
loop {
// Wait for the host info refresh interval.
// Sleep to calculate the network traffic difference over
// the DEFAULT_HOST_INFO_REFRESH_INTERVAL.
tokio::time::sleep(DEFAULT_HOST_INFO_REFRESH_INTERVAL).await;
// Wait for getting the network data.
let network_data = interface.get_network_data().await;
debug!(
"network data: rx bandwidth {}/{} bps, tx bandwidth {}/{} bps",
network_data.rx_bandwidth.unwrap_or(0),
network_data.max_rx_bandwidth,
network_data.tx_bandwidth.unwrap_or(0),
network_data.max_tx_bandwidth
);
// Refresh network information.
networks.refresh();
// Init response.
let mut host = Host::default();
if let Some(network_data) = networks.get(&interface.name) {
let network = Network {
download_rate: network_data.received()
/ DEFAULT_HOST_INFO_REFRESH_INTERVAL.as_secs(),
// Convert bandwidth to bytes per second.
download_rate_limit: interface.bandwidth / 8 * MB,
upload_rate: network_data.transmitted()
/ DEFAULT_HOST_INFO_REFRESH_INTERVAL.as_secs(),
// Convert bandwidth to bytes per second.
upload_rate_limit: interface.bandwidth / 8 * MB,
..Default::default()
};
host.network = Some(network.clone());
debug!("interface: {}, network: {:?}", interface.name, network);
};
// Send host info.
match out_stream_tx
.send(Ok(Host {
network: Some(Network {
max_rx_bandwidth: network_data.max_rx_bandwidth,
rx_bandwidth: network_data.rx_bandwidth,
max_tx_bandwidth: network_data.max_tx_bandwidth,
tx_bandwidth: network_data.tx_bandwidth,
..Default::default()
}),
..Default::default()
}))
.await
{
match out_stream_tx.send(Ok(host.clone())).await {
Ok(_) => {}
Err(err) => {
error!(
@ -1066,7 +1044,7 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
remote_host_id, err
);
return;
break;
}
};
}
@ -1082,7 +1060,7 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
ReceiverStream<Result<DownloadPersistentCacheTaskResponse, Status>>;
/// download_persistent_cache_task downloads the persistent cache task.
#[instrument(skip_all, fields(host_id, task_id, peer_id, remote_ip, content_length))]
#[instrument(skip_all, fields(host_id, task_id, peer_id))]
async fn download_persistent_cache_task(
&self,
request: Request<DownloadPersistentCacheTaskRequest>,
@ -1115,10 +1093,6 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
Span::current().record("host_id", host_id.as_str());
Span::current().record("task_id", task_id.as_str());
Span::current().record("peer_id", peer_id.as_str());
Span::current().record(
"remote_ip",
request.remote_ip.clone().unwrap_or_default().as_str(),
);
info!("download persistent cache task in download server");
// Download task started.
@ -1172,15 +1146,12 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
task
}
};
info!(
"content length {}, piece length {}",
task.content_length(),
task.piece_length()
);
Span::current().record("content_length", task.content_length());
// Initialize stream channel.
let request_clone = request.clone();
let task_manager_clone = self.persistent_cache_task.clone();
@ -1314,7 +1285,7 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
}
/// update_persistent_cache_task update metadata of the persistent cache task.
#[instrument(skip_all, fields(host_id, task_id, remote_ip))]
#[instrument(skip_all, fields(host_id, task_id))]
async fn update_persistent_cache_task(
&self,
request: Request<UpdatePersistentCacheTaskRequest>,
@ -1336,10 +1307,6 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
// Span record the host id and task id.
Span::current().record("host_id", host_id.as_str());
Span::current().record("task_id", task_id.as_str());
Span::current().record(
"remote_ip",
request.remote_ip.clone().unwrap_or_default().as_str(),
);
info!("update persistent cache task in upload server");
// Collect the update task started metrics.
@ -1361,7 +1328,7 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
}
/// stat_persistent_cache_task stats the persistent cache task.
#[instrument(skip_all, fields(host_id, task_id, remote_ip))]
#[instrument(skip_all, fields(host_id, task_id))]
async fn stat_persistent_cache_task(
&self,
request: Request<StatPersistentCacheTaskRequest>,
@ -1383,10 +1350,6 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
// Span record the host id and task id.
Span::current().record("host_id", host_id.as_str());
Span::current().record("task_id", task_id.as_str());
Span::current().record(
"remote_ip",
request.remote_ip.clone().unwrap_or_default().as_str(),
);
info!("stat persistent cache task in upload server");
// Collect the stat task started metrics.
@ -1408,7 +1371,7 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
}
/// delete_persistent_cache_task deletes the persistent cache task.
#[instrument(skip_all, fields(host_id, task_id, remote_ip))]
#[instrument(skip_all, fields(host_id, task_id))]
async fn delete_persistent_cache_task(
&self,
request: Request<DeletePersistentCacheTaskRequest>,
@ -1430,10 +1393,6 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
// Span record the host id and task id.
Span::current().record("host_id", host_id.as_str());
Span::current().record("task_id", task_id.as_str());
Span::current().record(
"remote_ip",
request.remote_ip.clone().unwrap_or_default().as_str(),
);
info!("delete persistent cache task in upload server");
// Collect the delete task started metrics.
@ -1486,6 +1445,7 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
tokio::spawn(
async move {
loop {
let mut has_started_piece = false;
let mut finished_piece_numbers = Vec::new();
for interested_piece_number in interested_piece_numbers.iter() {
let piece = match task_manager.piece.get(
@ -1544,6 +1504,11 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
finished_piece_numbers.push(piece.number);
continue;
}
// Check whether the piece is started.
if piece.is_started() {
has_started_piece = true;
}
}
// Remove the finished piece numbers from the interested piece numbers.
@ -1557,6 +1522,13 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
return;
}
// If there is no started piece, return.
if !has_started_piece {
info!("there is no started persistent cache piece");
drop(out_stream_tx);
return;
}
// Wait for the piece to be finished.
tokio::time::sleep(
dragonfly_client_storage::DEFAULT_WAIT_FOR_PIECE_FINISHED_INTERVAL,
@ -1658,7 +1630,6 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
error!("upload persistent cache piece content failed: {}", err);
Status::internal(err.to_string())
})?;
drop(reader);
// Collect upload piece finished metrics.
collect_upload_piece_finished_metrics();
@ -1691,63 +1662,6 @@ impl DfdaemonUpload for DfdaemonUploadServerHandler {
) -> Result<Response<ExchangeIbVerbsQueuePairEndpointResponse>, Status> {
unimplemented!()
}
/// DownloadCacheTaskStream is the stream of the download cache task response.
type DownloadCacheTaskStream = ReceiverStream<Result<DownloadCacheTaskResponse, Status>>;
/// download_cache_task downloads the cache task.
#[instrument(
skip_all,
fields(host_id, task_id, peer_id, url, remote_ip, content_length)
)]
async fn download_cache_task(
&self,
_request: Request<DownloadCacheTaskRequest>,
) -> Result<Response<Self::DownloadCacheTaskStream>, Status> {
todo!();
}
/// stat_cache_task stats the cache task.
#[instrument(skip_all, fields(host_id, task_id, remote_ip, local_only))]
async fn stat_cache_task(
&self,
_request: Request<StatCacheTaskRequest>,
) -> Result<Response<CacheTask>, Status> {
todo!();
}
/// delete_cache_task deletes the cache task.
#[instrument(skip_all, fields(host_id, task_id, remote_ip))]
async fn delete_cache_task(
&self,
_request: Request<DeleteCacheTaskRequest>,
) -> Result<Response<()>, Status> {
todo!();
}
/// SyncCachePiecesStream is the stream of the sync cache pieces response.
type SyncCachePiecesStream = ReceiverStream<Result<SyncCachePiecesResponse, Status>>;
/// sync_cache_pieces provides the cache piece metadata for parent.
#[instrument(skip_all, fields(host_id, remote_host_id, task_id))]
async fn sync_cache_pieces(
&self,
_request: Request<SyncCachePiecesRequest>,
) -> Result<Response<Self::SyncCachePiecesStream>, Status> {
todo!();
}
/// download_cache_piece provides the cache piece content for parent.
#[instrument(
skip_all,
fields(host_id, remote_host_id, task_id, piece_id, piece_length)
)]
async fn download_cache_piece(
&self,
_request: Request<DownloadCachePieceRequest>,
) -> Result<Response<DownloadCachePieceResponse>, Status> {
todo!();
}
}
/// DfdaemonUploadClient is a wrapper of DfdaemonUploadGRPCClient.
@ -1946,7 +1860,6 @@ impl DfdaemonUploadClient {
}
/// sync_persistent_cache_pieces provides the persistent cache piece metadata for parent.
/// If the per-piece collection timeout is exceeded, the stream will be closed.
#[instrument(skip_all)]
pub async fn sync_persistent_cache_pieces(
&self,

View File

@ -50,11 +50,11 @@ pub const HTTP2_KEEP_ALIVE_INTERVAL: Duration = Duration::from_secs(300);
/// HTTP2_KEEP_ALIVE_TIMEOUT is the timeout for HTTP2 keep alive.
pub const HTTP2_KEEP_ALIVE_TIMEOUT: Duration = Duration::from_secs(20);
/// MAX_FRAME_SIZE is the max frame size for GRPC, default is 4MB.
pub const MAX_FRAME_SIZE: u32 = 4 * 1024 * 1024;
/// MAX_FRAME_SIZE is the max frame size for GRPC, default is 12MB.
pub const MAX_FRAME_SIZE: u32 = 12 * 1024 * 1024;
/// INITIAL_WINDOW_SIZE is the initial window size for GRPC, default is 512KB.
pub const INITIAL_WINDOW_SIZE: u32 = 512 * 1024;
/// INITIAL_WINDOW_SIZE is the initial window size for GRPC, default is 12MB.
pub const INITIAL_WINDOW_SIZE: u32 = 12 * 1024 * 1024;
/// BUFFER_SIZE is the buffer size for GRPC, default is 64KB.
pub const BUFFER_SIZE: usize = 64 * 1024;

View File

@ -26,8 +26,9 @@ use prometheus::{
};
use std::net::SocketAddr;
use std::path::Path;
use std::sync::Arc;
use std::sync::{Arc, Mutex};
use std::time::Duration;
use sysinfo::{ProcessRefreshKind, ProcessesToUpdate, RefreshKind, System, UpdateKind};
use tokio::sync::mpsc;
use tracing::{error, info, instrument, warn};
use warp::{Filter, Rejection, Reply};
@ -212,21 +213,7 @@ lazy_static! {
&["type"]
).expect("metric can be created");
/// LIST_TASK_ENTRIES_COUNT is used to count the number of list task entries.
pub static ref LIST_TASK_ENTRIES_COUNT: IntCounterVec =
IntCounterVec::new(
Opts::new("list_task_entries_total", "Counter of the number of the list task entries.").namespace(dragonfly_client_config::SERVICE_NAME).subsystem(dragonfly_client_config::NAME),
&["type"]
).expect("metric can be created");
/// LIST_TASK_ENTRIES_FAILURE_COUNT is used to count the failed number of list task entries.
pub static ref LIST_TASK_ENTRIES_FAILURE_COUNT: IntCounterVec =
IntCounterVec::new(
Opts::new("list_task_entries_failure_total", "Counter of the number of failed of the list task entries.").namespace(dragonfly_client_config::SERVICE_NAME).subsystem(dragonfly_client_config::NAME),
&["type"]
).expect("metric can be created");
/// DELETE_TASK_COUNT is used to count the number of delete tasks.
/// DELETE_TASK_COUNT is used to count the number of delete tasks.
pub static ref DELETE_TASK_COUNT: IntCounterVec =
IntCounterVec::new(
Opts::new("delete_task_total", "Counter of the number of the delete task.").namespace(dragonfly_client_config::SERVICE_NAME).subsystem(dragonfly_client_config::NAME),
@ -267,6 +254,20 @@ lazy_static! {
Opts::new("disk_usage_space_total", "Gauge of the disk usage space in bytes").namespace(dragonfly_client_config::SERVICE_NAME).subsystem(dragonfly_client_config::NAME),
&[]
).expect("metric can be created");
/// DISK_WRITTEN_BYTES is used to count of the disk written bytes.
pub static ref DISK_WRITTEN_BYTES: IntGaugeVec =
IntGaugeVec::new(
Opts::new("disk_written_bytes", "Gauge of the disk written bytes.").namespace(dragonfly_client_config::SERVICE_NAME).subsystem(dragonfly_client_config::NAME),
&[]
).expect("metric can be created");
/// DISK_READ_BYTES is used to count of the disk read bytes.
pub static ref DISK_READ_BYTES: IntGaugeVec =
IntGaugeVec::new(
Opts::new("disk_read_bytes", "Gauge of the disk read bytes.").namespace(dragonfly_client_config::SERVICE_NAME).subsystem(dragonfly_client_config::NAME),
&[]
).expect("metric can be created");
}
/// register_custom_metrics registers all custom metrics.
@ -351,14 +352,6 @@ fn register_custom_metrics() {
.register(Box::new(STAT_TASK_FAILURE_COUNT.clone()))
.expect("metric can be registered");
REGISTRY
.register(Box::new(LIST_TASK_ENTRIES_COUNT.clone()))
.expect("metric can be registered");
REGISTRY
.register(Box::new(LIST_TASK_ENTRIES_FAILURE_COUNT.clone()))
.expect("metric can be registered");
REGISTRY
.register(Box::new(DELETE_TASK_COUNT.clone()))
.expect("metric can be registered");
@ -382,6 +375,14 @@ fn register_custom_metrics() {
REGISTRY
.register(Box::new(DISK_USAGE_SPACE.clone()))
.expect("metric can be registered");
REGISTRY
.register(Box::new(DISK_WRITTEN_BYTES.clone()))
.expect("metric can be registered");
REGISTRY
.register(Box::new(DISK_READ_BYTES.clone()))
.expect("metric can be registered");
}
/// reset_custom_metrics resets all custom metrics.
@ -406,14 +407,14 @@ fn reset_custom_metrics() {
UPDATE_TASK_FAILURE_COUNT.reset();
STAT_TASK_COUNT.reset();
STAT_TASK_FAILURE_COUNT.reset();
LIST_TASK_ENTRIES_COUNT.reset();
LIST_TASK_ENTRIES_FAILURE_COUNT.reset();
DELETE_TASK_COUNT.reset();
DELETE_TASK_FAILURE_COUNT.reset();
DELETE_HOST_COUNT.reset();
DELETE_HOST_FAILURE_COUNT.reset();
DISK_SPACE.reset();
DISK_USAGE_SPACE.reset();
DISK_WRITTEN_BYTES.reset();
DISK_READ_BYTES.reset();
}
/// TaskSize represents the size of the task.
@ -772,20 +773,6 @@ pub fn collect_stat_task_failure_metrics(typ: i32) {
.inc();
}
/// collect_list_task_entries_started_metrics collects the list task entries started metrics.
pub fn collect_list_task_entries_started_metrics(typ: i32) {
LIST_TASK_ENTRIES_COUNT
.with_label_values(&[typ.to_string().as_str()])
.inc();
}
/// collect_list_task_entries_failure_metrics collects the list task entries failure metrics.
pub fn collect_list_task_entries_failure_metrics(typ: i32) {
LIST_TASK_ENTRIES_FAILURE_COUNT
.with_label_values(&[typ.to_string().as_str()])
.inc();
}
/// collect_delete_task_started_metrics collects the delete task started metrics.
pub fn collect_delete_task_started_metrics(typ: i32) {
DELETE_TASK_COUNT
@ -811,7 +798,7 @@ pub fn collect_delete_host_failure_metrics() {
}
/// collect_disk_metrics collects the disk metrics.
pub fn collect_disk_metrics(path: &Path) {
pub fn collect_disk_metrics(path: &Path, system: &Arc<Mutex<System>>) {
// Collect disk space metrics.
let stats = match fs2::statvfs(path) {
Ok(stats) => stats,
@ -828,6 +815,24 @@ pub fn collect_disk_metrics(path: &Path) {
DISK_USAGE_SPACE
.with_label_values(&[])
.set(usage_space as i64);
// Collect disk bandwidth metrics.
let mut sys = system.lock().unwrap();
sys.refresh_processes_specifics(
ProcessesToUpdate::All,
true,
ProcessRefreshKind::new()
.with_disk_usage()
.with_exe(UpdateKind::Always),
);
let process = sys.process(sysinfo::get_current_pid().unwrap()).unwrap();
DISK_WRITTEN_BYTES
.with_label_values(&[])
.set(process.disk_usage().written_bytes as i64);
DISK_READ_BYTES
.with_label_values(&[])
.set(process.disk_usage().read_bytes as i64);
}
/// Metrics is the metrics server.
@ -836,6 +841,9 @@ pub struct Metrics {
/// config is the configuration of the dfdaemon.
config: Arc<Config>,
// system is the system information, only used for collecting disk metrics.
system: Arc<Mutex<System>>,
/// shutdown is used to shutdown the metrics server.
shutdown: shutdown::Shutdown,
@ -853,6 +861,13 @@ impl Metrics {
) -> Self {
Self {
config,
system: Arc::new(Mutex::new(System::new_with_specifics(
RefreshKind::new().with_processes(
ProcessRefreshKind::new()
.with_disk_usage()
.with_exe(UpdateKind::Always),
),
))),
shutdown,
_shutdown_complete: shutdown_complete_tx,
}
@ -879,6 +894,7 @@ impl Metrics {
// Clone the config.
let config = self.config.clone();
let system = self.system.clone();
// Create the metrics server address.
let addr = SocketAddr::new(
@ -890,7 +906,7 @@ impl Metrics {
let get_metrics_route = warp::path!("metrics")
.and(warp::get())
.and(warp::path::end())
.and_then(move || Self::get_metrics_handler(config.clone()));
.and_then(move || Self::get_metrics_handler(config.clone(), system.clone()));
// Delete the metrics route.
let delete_metrics_route = warp::path!("metrics")
@ -915,9 +931,12 @@ impl Metrics {
/// get_metrics_handler handles the metrics request of getting.
#[instrument(skip_all)]
async fn get_metrics_handler(config: Arc<Config>) -> Result<impl Reply, Rejection> {
async fn get_metrics_handler(
config: Arc<Config>,
system: Arc<Mutex<System>>,
) -> Result<impl Reply, Rejection> {
// Collect the disk space metrics.
collect_disk_metrics(config.storage.dir.as_path());
collect_disk_metrics(config.storage.dir.as_path(), &system);
// Encode custom metrics.
let encoder = TextEncoder::new();

View File

@ -66,28 +66,19 @@ pub const DRAGONFLY_OUTPUT_PATH_HEADER: &str = "X-Dragonfly-Output-Path";
/// For more details refer to https://github.com/dragonflyoss/design/blob/main/systems-analysis/file-download-workflow-with-hard-link/README.md.
pub const DRAGONFLY_FORCE_HARD_LINK_HEADER: &str = "X-Dragonfly-Force-Hard-Link";
/// DRAGONFLY_PIECE_LENGTH_HEADER is the header key of piece length in http request.
/// DRAGONFLY_PIECE_LENGTH is the header key of piece length in http request.
/// If the value is set, the piece length will be used to download the file.
/// Different piece length will generate different task id. The value needs to
/// be set with human readable format and needs to be greater than or equal
/// to 4mib, for example: 4mib, 1gib
pub const DRAGONFLY_PIECE_LENGTH_HEADER: &str = "X-Dragonfly-Piece-Length";
pub const DRAGONFLY_PIECE_LENGTH: &str = "X-Dragonfly-Piece-Length";
/// DRAGONFLY_CONTENT_FOR_CALCULATING_TASK_ID_HEADER is the header key of content for calculating task id.
/// If DRAGONFLY_CONTENT_FOR_CALCULATING_TASK_ID_HEADER is set, use its value to calculate the task ID.
/// DRAGONFLY_CONTENT_FOR_CALCULATING_TASK_ID is the header key of content for calculating task id.
/// If DRAGONFLY_CONTENT_FOR_CALCULATING_TASK_ID is set, use its value to calculate the task ID.
/// Otherwise, calculate the task ID based on `url`, `piece_length`, `tag`, `application`, and `filtered_query_params`.
pub const DRAGONFLY_CONTENT_FOR_CALCULATING_TASK_ID_HEADER: &str =
pub const DRAGONFLY_CONTENT_FOR_CALCULATING_TASK_ID: &str =
"X-Dragonfly-Content-For-Calculating-Task-ID";
/// DRAGONFLY_TASK_DOWNLOAD_FINISHED_HEADER is the response header key to indicate whether the task download finished.
/// When the task download is finished, the response will include this header with the value `"true"`,
/// indicating that the download hit the local cache.
pub const DRAGONFLY_TASK_DOWNLOAD_FINISHED_HEADER: &str = "X-Dragonfly-Task-Download-Finished";
/// DRAGONFLY_TASK_ID_HEADER is the response header key of task id. Client will calculate the task ID
/// based on `url`, `piece_length`, `tag`, `application`, and `filtered_query_params`.
pub const DRAGONFLY_TASK_ID_HEADER: &str = "X-Dragonfly-Task-ID";
/// get_tag gets the tag from http header.
pub fn get_tag(header: &HeaderMap) -> Option<String> {
header
@ -202,7 +193,7 @@ pub fn get_force_hard_link(header: &HeaderMap) -> bool {
/// get_piece_length gets the piece length from http header.
pub fn get_piece_length(header: &HeaderMap) -> Option<ByteSize> {
match header.get(DRAGONFLY_PIECE_LENGTH_HEADER) {
match header.get(DRAGONFLY_PIECE_LENGTH) {
Some(piece_length) => match piece_length.to_str() {
Ok(piece_length) => match piece_length.parse::<ByteSize>() {
Ok(piece_length) => Some(piece_length),
@ -223,7 +214,7 @@ pub fn get_piece_length(header: &HeaderMap) -> Option<ByteSize> {
/// get_content_for_calculating_task_id gets the content for calculating task id from http header.
pub fn get_content_for_calculating_task_id(header: &HeaderMap) -> Option<String> {
header
.get(DRAGONFLY_CONTENT_FOR_CALCULATING_TASK_ID_HEADER)
.get(DRAGONFLY_CONTENT_FOR_CALCULATING_TASK_ID)
.and_then(|content| content.to_str().ok())
.map(|content| content.to_string())
}
@ -368,22 +359,16 @@ mod tests {
#[test]
fn test_get_piece_length() {
let mut headers = HeaderMap::new();
headers.insert(
DRAGONFLY_PIECE_LENGTH_HEADER,
HeaderValue::from_static("4mib"),
);
headers.insert(DRAGONFLY_PIECE_LENGTH, HeaderValue::from_static("4mib"));
assert_eq!(get_piece_length(&headers), Some(ByteSize::mib(4)));
let empty_headers = HeaderMap::new();
assert_eq!(get_piece_length(&empty_headers), None);
headers.insert(
DRAGONFLY_PIECE_LENGTH_HEADER,
HeaderValue::from_static("invalid"),
);
headers.insert(DRAGONFLY_PIECE_LENGTH, HeaderValue::from_static("invalid"));
assert_eq!(get_piece_length(&headers), None);
headers.insert(DRAGONFLY_PIECE_LENGTH_HEADER, HeaderValue::from_static("0"));
headers.insert(DRAGONFLY_PIECE_LENGTH, HeaderValue::from_static("0"));
assert_eq!(get_piece_length(&headers), Some(ByteSize::b(0)));
}
@ -391,7 +376,7 @@ mod tests {
fn test_get_content_for_calculating_task_id() {
let mut headers = HeaderMap::new();
headers.insert(
DRAGONFLY_CONTENT_FOR_CALCULATING_TASK_ID_HEADER,
DRAGONFLY_CONTENT_FOR_CALCULATING_TASK_ID,
HeaderValue::from_static("test-content"),
);
assert_eq!(

View File

@ -208,7 +208,7 @@ impl Proxy {
service_fn(move |request|{
let context = context.clone();
async move {
handler(context.config, context.task, request, context.dfdaemon_download_client, context.registry_cert, context.server_ca_cert, remote_address.ip()).await
handler(context.config, context.task, request, context.dfdaemon_download_client, context.registry_cert, context.server_ca_cert).await
}
} ),
)
@ -231,7 +231,7 @@ impl Proxy {
}
/// handler handles the request from the client.
#[instrument(skip_all, fields(url, method, remote_ip))]
#[instrument(skip_all, fields(uri, method))]
pub async fn handler(
config: Arc<Config>,
task: Arc<Task>,
@ -239,13 +239,7 @@ pub async fn handler(
dfdaemon_download_client: DfdaemonDownloadClient,
registry_cert: Arc<Option<Vec<CertificateDer<'static>>>>,
server_ca_cert: Arc<Option<Certificate>>,
remote_ip: std::net::IpAddr,
) -> ClientResult<Response> {
// Span record the url and method.
Span::current().record("url", request.uri().to_string().as_str());
Span::current().record("method", request.method().as_str());
Span::current().record("remote_ip", remote_ip.to_string().as_str());
// Record the proxy request started metrics. The metrics will be recorded
// when the request is kept alive.
collect_proxy_request_started_metrics();
@ -258,7 +252,6 @@ pub async fn handler(
config,
task,
request,
remote_ip,
dfdaemon_download_client,
registry_cert,
server_ca_cert,
@ -270,20 +263,22 @@ pub async fn handler(
config,
task,
request,
remote_ip,
dfdaemon_download_client,
registry_cert,
)
.await;
}
// Span record the uri and method.
Span::current().record("uri", request.uri().to_string().as_str());
Span::current().record("method", request.method().as_str());
// Handle CONNECT request.
if Method::CONNECT == request.method() {
return https_handler(
config,
task,
request,
remote_ip,
dfdaemon_download_client,
registry_cert,
server_ca_cert,
@ -295,7 +290,6 @@ pub async fn handler(
config,
task,
request,
remote_ip,
dfdaemon_download_client,
registry_cert,
)
@ -308,7 +302,6 @@ pub async fn registry_mirror_http_handler(
config: Arc<Config>,
task: Arc<Task>,
request: Request<hyper::body::Incoming>,
remote_ip: std::net::IpAddr,
dfdaemon_download_client: DfdaemonDownloadClient,
registry_cert: Arc<Option<Vec<CertificateDer<'static>>>>,
) -> ClientResult<Response> {
@ -317,7 +310,6 @@ pub async fn registry_mirror_http_handler(
config,
task,
request,
remote_ip,
dfdaemon_download_client,
registry_cert,
)
@ -330,7 +322,6 @@ pub async fn registry_mirror_https_handler(
config: Arc<Config>,
task: Arc<Task>,
request: Request<hyper::body::Incoming>,
remote_ip: std::net::IpAddr,
dfdaemon_download_client: DfdaemonDownloadClient,
registry_cert: Arc<Option<Vec<CertificateDer<'static>>>>,
server_ca_cert: Arc<Option<Certificate>>,
@ -340,7 +331,6 @@ pub async fn registry_mirror_https_handler(
config,
task,
request,
remote_ip,
dfdaemon_download_client,
registry_cert,
server_ca_cert,
@ -354,7 +344,6 @@ pub async fn http_handler(
config: Arc<Config>,
task: Arc<Task>,
request: Request<hyper::body::Incoming>,
remote_ip: std::net::IpAddr,
dfdaemon_download_client: DfdaemonDownloadClient,
registry_cert: Arc<Option<Vec<CertificateDer<'static>>>>,
) -> ClientResult<Response> {
@ -386,15 +375,7 @@ pub async fn http_handler(
request.method(),
request_uri
);
return proxy_via_dfdaemon(
config,
task,
&rule,
request,
remote_ip,
dfdaemon_download_client,
)
.await;
return proxy_via_dfdaemon(config, task, &rule, request, dfdaemon_download_client).await;
}
// If the request header contains the X-Dragonfly-Use-P2P header, proxy the request via the
@ -410,7 +391,6 @@ pub async fn http_handler(
task,
&Rule::default(),
request,
remote_ip,
dfdaemon_download_client,
)
.await;
@ -439,7 +419,6 @@ pub async fn https_handler(
config: Arc<Config>,
task: Arc<Task>,
request: Request<hyper::body::Incoming>,
remote_ip: std::net::IpAddr,
dfdaemon_download_client: DfdaemonDownloadClient,
registry_cert: Arc<Option<Vec<CertificateDer<'static>>>>,
server_ca_cert: Arc<Option<Certificate>>,
@ -459,7 +438,6 @@ pub async fn https_handler(
upgraded,
host,
port,
remote_ip,
dfdaemon_download_client,
registry_cert,
server_ca_cert,
@ -490,7 +468,6 @@ async fn upgraded_tunnel(
upgraded: Upgraded,
host: String,
port: u16,
remote_ip: std::net::IpAddr,
dfdaemon_download_client: DfdaemonDownloadClient,
registry_cert: Arc<Option<Vec<CertificateDer<'static>>>>,
server_ca_cert: Arc<Option<Certificate>>,
@ -539,7 +516,6 @@ async fn upgraded_tunnel(
host.clone(),
port,
request,
remote_ip,
dfdaemon_download_client.clone(),
registry_cert.clone(),
)
@ -555,20 +531,18 @@ async fn upgraded_tunnel(
}
/// upgraded_handler handles the upgraded https request from the client.
#[allow(clippy::too_many_arguments)]
#[instrument(skip_all, fields(url, method))]
#[instrument(skip_all, fields(uri, method))]
pub async fn upgraded_handler(
config: Arc<Config>,
task: Arc<Task>,
host: String,
port: u16,
mut request: Request<hyper::body::Incoming>,
remote_ip: std::net::IpAddr,
dfdaemon_download_client: DfdaemonDownloadClient,
registry_cert: Arc<Option<Vec<CertificateDer<'static>>>>,
) -> ClientResult<Response> {
// Span record the url and method.
Span::current().record("url", request.uri().to_string().as_str());
// Span record the uri and method.
Span::current().record("uri", request.uri().to_string().as_str());
Span::current().record("method", request.method().as_str());
// Authenticate the request with the basic auth.
@ -613,15 +587,7 @@ pub async fn upgraded_handler(
request.method(),
request_uri
);
return proxy_via_dfdaemon(
config,
task,
&rule,
request,
remote_ip,
dfdaemon_download_client,
)
.await;
return proxy_via_dfdaemon(config, task, &rule, request, dfdaemon_download_client).await;
}
// If the request header contains the X-Dragonfly-Use-P2P header, proxy the request via the
@ -637,7 +603,6 @@ pub async fn upgraded_handler(
task,
&Rule::default(),
request,
remote_ip,
dfdaemon_download_client,
)
.await;
@ -667,24 +632,22 @@ async fn proxy_via_dfdaemon(
task: Arc<Task>,
rule: &Rule,
request: Request<hyper::body::Incoming>,
remote_ip: std::net::IpAddr,
dfdaemon_download_client: DfdaemonDownloadClient,
) -> ClientResult<Response> {
// Collect the metrics for the proxy request via dfdaemon.
collect_proxy_request_via_dfdaemon_metrics();
// Make the download task request.
let download_task_request =
match make_download_task_request(config.clone(), rule, request, remote_ip) {
Ok(download_task_request) => download_task_request,
Err(err) => {
error!("make download task request failed: {}", err);
return Ok(make_error_response(
http::StatusCode::INTERNAL_SERVER_ERROR,
None,
));
}
};
let download_task_request = match make_download_task_request(config.clone(), rule, request) {
Ok(download_task_request) => download_task_request,
Err(err) => {
error!("make download task request failed: {}", err);
return Ok(make_error_response(
http::StatusCode::INTERNAL_SERVER_ERROR,
None,
));
}
};
// Download the task by the dfdaemon download client.
let response = match dfdaemon_download_client
@ -768,10 +731,7 @@ async fn proxy_via_dfdaemon(
// Construct the response.
let mut response = Response::new(boxed_body);
*response.headers_mut() = make_response_headers(
message.task_id.as_str(),
download_task_started_response.clone(),
)?;
*response.headers_mut() = make_response_headers(download_task_started_response.clone())?;
*response.status_mut() = http::StatusCode::OK;
// Return the response if the client return the first piece.
@ -1056,7 +1016,6 @@ fn make_download_task_request(
config: Arc<Config>,
rule: &Rule,
request: Request<hyper::body::Incoming>,
remote_ip: std::net::IpAddr,
) -> ClientResult<DownloadTaskRequest> {
// Convert the Reqwest header to the Hyper header.
let mut header = request.headers().clone();
@ -1102,9 +1061,9 @@ fn make_download_task_request(
hdfs: None,
is_prefetch: false,
need_piece_content: false,
load_to_cache: false,
force_hard_link: header::get_force_hard_link(&header),
content_for_calculating_task_id: header::get_content_for_calculating_task_id(&header),
remote_ip: Some(remote_ip.to_string()),
}),
})
}
@ -1153,7 +1112,6 @@ fn make_download_url(
/// make_response_headers makes the response headers.
fn make_response_headers(
task_id: &str,
mut download_task_started_response: DownloadTaskStartedResponse,
) -> ClientResult<hyper::header::HeaderMap> {
// Insert the content range header to the response header.
@ -1174,18 +1132,6 @@ fn make_response_headers(
);
};
if download_task_started_response.is_finished {
download_task_started_response.response_header.insert(
header::DRAGONFLY_TASK_DOWNLOAD_FINISHED_HEADER.to_string(),
"true".to_string(),
);
}
download_task_started_response.response_header.insert(
header::DRAGONFLY_TASK_ID_HEADER.to_string(),
task_id.to_string(),
);
hashmap_to_headermap(&download_task_started_response.response_header)
}

View File

@ -140,15 +140,11 @@ impl Piece {
}
/// get gets a piece from the local storage.
#[instrument(skip_all)]
pub fn get(&self, piece_id: &str) -> Result<Option<metadata::Piece>> {
self.storage.get_piece(piece_id)
}
/// get_all gets all pieces of a task from the local storage.
pub fn get_all(&self, task_id: &str) -> Result<Vec<metadata::Piece>> {
self.storage.get_pieces(task_id)
}
/// calculate_interested calculates the interested pieces by content_length and range.
pub fn calculate_interested(
&self,
@ -340,7 +336,6 @@ impl Piece {
) -> Result<impl AsyncRead> {
// Span record the piece_id.
Span::current().record("piece_id", piece_id);
Span::current().record("piece_length", length);
// Acquire the upload rate limiter.
if !disable_rate_limit {
@ -372,7 +367,6 @@ impl Piece {
) -> Result<impl AsyncRead> {
// Span record the piece_id.
Span::current().record("piece_id", piece_id);
Span::current().record("piece_length", length);
// Acquire the download rate limiter.
if !disable_rate_limit {
@ -412,10 +406,10 @@ impl Piece {
length: u64,
parent: piece_collector::CollectedParent,
is_prefetch: bool,
load_to_cache: bool,
) -> Result<metadata::Piece> {
// Span record the piece_id.
Span::current().record("piece_id", piece_id);
Span::current().record("piece_length", length);
// Record the start of downloading piece.
let piece = self
@ -476,6 +470,7 @@ impl Piece {
digest.as_str(),
parent.id.as_str(),
&mut reader,
load_to_cache,
self.config.storage.write_piece_timeout,
)
.await
@ -513,12 +508,12 @@ impl Piece {
length: u64,
request_header: HeaderMap,
is_prefetch: bool,
load_to_cache: bool,
object_storage: Option<ObjectStorage>,
hdfs: Option<Hdfs>,
) -> Result<metadata::Piece> {
// Span record the piece_id.
Span::current().record("piece_id", piece_id);
Span::current().record("piece_length", length);
// Record the start of downloading piece.
let piece = self
@ -638,6 +633,7 @@ impl Piece {
offset,
length,
&mut response.reader,
load_to_cache,
self.config.storage.write_piece_timeout,
)
.await
@ -701,7 +697,6 @@ impl Piece {
) -> Result<impl AsyncRead> {
// Span record the piece_id.
Span::current().record("piece_id", piece_id);
Span::current().record("piece_length", length);
// Acquire the upload rate limiter.
self.upload_rate_limiter.acquire(length as usize).await;
@ -731,7 +726,6 @@ impl Piece {
) -> Result<impl AsyncRead> {
// Span record the piece_id.
Span::current().record("piece_id", piece_id);
Span::current().record("piece_length", length);
// Acquire the download rate limiter.
if !disable_rate_limit {
@ -776,7 +770,6 @@ impl Piece {
) -> Result<metadata::Piece> {
// Span record the piece_id.
Span::current().record("piece_id", piece_id);
Span::current().record("piece_length", length);
if is_prefetch {
// Acquire the prefetch rate limiter.
@ -841,7 +834,6 @@ impl Piece {
piece_id,
task_id,
offset,
length,
digest.as_str(),
parent.id.as_str(),
&mut reader,

View File

@ -15,21 +15,20 @@
*/
use crate::grpc::dfdaemon_upload::DfdaemonUploadClient;
use dashmap::DashMap;
use dragonfly_api::common::v2::Host;
use dragonfly_api::dfdaemon::v2::{SyncPersistentCachePiecesRequest, SyncPiecesRequest};
use dragonfly_client_config::dfdaemon::Config;
use dragonfly_client_core::{Error, Result};
use dragonfly_client_storage::metadata;
use std::collections::HashMap;
use std::sync::Arc;
use std::time::Duration;
use tokio::sync::mpsc::{self, Receiver, Sender};
use tokio::sync::Mutex;
use tokio::task::JoinSet;
use tokio_stream::StreamExt;
use tracing::{error, info, instrument, Instrument};
const DEFAULT_WAIT_FOR_PIECE_FROM_DIFFERENT_PARENTS: Duration = Duration::from_millis(5);
/// CollectedParent is the parent peer collected from the parent.
#[derive(Clone, Debug)]
pub struct CollectedParent {
@ -69,8 +68,8 @@ pub struct PieceCollector {
/// interested_pieces is the pieces interested by the collector.
interested_pieces: Vec<metadata::Piece>,
/// collected_pieces is a map to store the collected pieces from different parents.
collected_pieces: Arc<DashMap<u32, Vec<CollectedParent>>>,
/// collected_pieces is the pieces collected from peers.
collected_pieces: Arc<Mutex<HashMap<u32, String>>>,
}
/// PieceCollector is used to collect pieces from peers.
@ -83,10 +82,14 @@ impl PieceCollector {
interested_pieces: Vec<metadata::Piece>,
parents: Vec<CollectedParent>,
) -> Self {
let collected_pieces = Arc::new(DashMap::with_capacity(interested_pieces.len()));
let collected_pieces =
Arc::new(Mutex::new(HashMap::with_capacity(interested_pieces.len())));
let mut collected_pieces_guard = collected_pieces.lock().await;
for interested_piece in &interested_pieces {
collected_pieces.insert(interested_piece.number, Vec::new());
collected_pieces_guard.insert(interested_piece.number, String::new());
}
drop(collected_pieces_guard);
Self {
config,
@ -107,7 +110,7 @@ impl PieceCollector {
let parents = self.parents.clone();
let interested_pieces = self.interested_pieces.clone();
let collected_pieces = self.collected_pieces.clone();
let collected_piece_timeout = self.config.download.collected_piece_timeout;
let collected_piece_timeout = self.config.download.piece_timeout;
let (collected_piece_tx, collected_piece_rx) = mpsc::channel(128 * 1024);
tokio::spawn(
async move {
@ -132,25 +135,7 @@ impl PieceCollector {
collected_piece_rx
}
/// collect_from_parents collects pieces from multiple parents with load balancing strategy.
///
/// The collection process works in two phases:
/// 1. **Synchronization Phase**: Waits for a configured duration (DEFAULT_WAIT_FOR_PIECE_FROM_DIFFERENT_PARENTS)
/// to collect the same piece information from different parents. This allows the collector
/// to gather multiple sources for each piece.
///
/// 2. **Selection Phase**: After the wait period, randomly selects one parent from the available
/// candidates for each piece and forwards it to the piece downloader.
///
/// **Load Balancing Strategy**:
/// The random parent selection is designed to distribute download load across multiple parents
/// during concurrent piece downloads. This approach ensures:
/// - Optimal utilization of bandwidth from multiple parent nodes
/// - Prevention of overwhelming any single parent with too many requests
/// - Better overall download performance through parallel connections
///
/// This strategy is particularly effective when downloading multiple pieces simultaneously,
/// as it naturally spreads the workload across the available parent pool.
/// collect_from_parents collects pieces from parents.
#[allow(clippy::too_many_arguments)]
#[instrument(skip_all)]
async fn collect_from_parents(
@ -159,7 +144,7 @@ impl PieceCollector {
task_id: &str,
parents: Vec<CollectedParent>,
interested_pieces: Vec<metadata::Piece>,
collected_pieces: Arc<DashMap<u32, Vec<CollectedParent>>>,
collected_pieces: Arc<Mutex<HashMap<u32, String>>>,
collected_piece_tx: Sender<CollectedPiece>,
collected_piece_timeout: Duration,
) -> Result<()> {
@ -173,7 +158,7 @@ impl PieceCollector {
task_id: String,
parent: CollectedParent,
interested_pieces: Vec<metadata::Piece>,
collected_pieces: Arc<DashMap<u32, Vec<CollectedParent>>>,
collected_pieces: Arc<Mutex<HashMap<u32, String>>>,
collected_piece_tx: Sender<CollectedPiece>,
collected_piece_timeout: Duration,
) -> Result<CollectedParent> {
@ -221,33 +206,18 @@ impl PieceCollector {
error!("sync pieces from parent {} failed: {}", parent.id, err);
})? {
let message = message?;
if let Some(mut parents) = collected_pieces.get_mut(&message.number) {
parents.push(parent.clone());
} else {
continue;
}
// Wait for collecting the piece from different parents when the first
// piece is collected.
tokio::time::sleep(DEFAULT_WAIT_FOR_PIECE_FROM_DIFFERENT_PARENTS).await;
let parents = match collected_pieces.remove(&message.number) {
Some((_, parents)) => parents,
None => continue,
};
let parent = match parents.get(fastrand::usize(..parents.len())) {
Some(parent) => parent,
None => {
error!(
"collected_pieces does not contain parent for piece {}",
message.number
);
// Remove the piece from collected_pieces, avoid to collect the same piece from
// different parents.
{
let mut collected_pieces_guard = collected_pieces.lock().await;
if collected_pieces_guard.remove(&message.number).is_none() {
continue;
}
};
}
info!(
"picked up piece {}-{} metadata from parent {}",
"received piece {}-{} metadata from parent {}",
task_id, message.number, parent.id
);
@ -288,7 +258,11 @@ impl PieceCollector {
info!("peer {} sync pieces finished", peer.id);
// If all pieces are collected, abort all tasks.
if collected_pieces.is_empty() {
let collected_pieces_guard = collected_pieces.lock().await;
let is_empty = collected_pieces_guard.is_empty();
drop(collected_pieces_guard);
if is_empty {
info!("all pieces are collected, abort all tasks");
join_set.abort_all();
}
@ -323,8 +297,8 @@ pub struct PersistentCachePieceCollector {
/// interested_pieces is the pieces interested by the collector.
interested_pieces: Vec<metadata::Piece>,
/// collected_pieces is a map to store the collected pieces from different parents.
collected_pieces: Arc<DashMap<u32, Vec<CollectedParent>>>,
/// collected_pieces is the pieces collected from peers.
collected_pieces: Arc<Mutex<HashMap<u32, String>>>,
}
/// PersistentCachePieceCollector is used to collect persistent cache pieces from peers.
@ -337,10 +311,14 @@ impl PersistentCachePieceCollector {
interested_pieces: Vec<metadata::Piece>,
parents: Vec<CollectedParent>,
) -> Self {
let collected_pieces = Arc::new(DashMap::with_capacity(interested_pieces.len()));
let collected_pieces =
Arc::new(Mutex::new(HashMap::with_capacity(interested_pieces.len())));
let mut collected_pieces_guard = collected_pieces.lock().await;
for interested_piece in &interested_pieces {
collected_pieces.insert(interested_piece.number, Vec::new());
collected_pieces_guard.insert(interested_piece.number, String::new());
}
drop(collected_pieces_guard);
Self {
config,
@ -386,25 +364,7 @@ impl PersistentCachePieceCollector {
collected_piece_rx
}
/// collect_from_parents collects pieces from multiple parents with load balancing strategy.
///
/// The collection process works in two phases:
/// 1. **Synchronization Phase**: Waits for a configured duration (DEFAULT_WAIT_FOR_PIECE_FROM_DIFFERENT_PARENTS)
/// to collect the same piece information from different parents. This allows the collector
/// to gather multiple sources for each piece.
///
/// 2. **Selection Phase**: After the wait period, randomly selects one parent from the available
/// candidates for each piece and forwards it to the piece downloader.
///
/// **Load Balancing Strategy**:
/// The random parent selection is designed to distribute download load across multiple parents
/// during concurrent piece downloads. This approach ensures:
/// - Optimal utilization of bandwidth from multiple parent nodes
/// - Prevention of overwhelming any single parent with too many requests
/// - Better overall download performance through parallel connections
///
/// This strategy is particularly effective when downloading multiple pieces simultaneously,
/// as it naturally spreads the workload across the available parent pool.
/// collect_from_parents collects pieces from parents.
#[allow(clippy::too_many_arguments)]
#[instrument(skip_all)]
async fn collect_from_parents(
@ -413,7 +373,7 @@ impl PersistentCachePieceCollector {
task_id: &str,
parents: Vec<CollectedParent>,
interested_pieces: Vec<metadata::Piece>,
collected_pieces: Arc<DashMap<u32, Vec<CollectedParent>>>,
collected_pieces: Arc<Mutex<HashMap<u32, String>>>,
collected_piece_tx: Sender<CollectedPiece>,
collected_piece_timeout: Duration,
) -> Result<()> {
@ -427,7 +387,7 @@ impl PersistentCachePieceCollector {
task_id: String,
parent: CollectedParent,
interested_pieces: Vec<metadata::Piece>,
collected_pieces: Arc<DashMap<u32, Vec<CollectedParent>>>,
collected_pieces: Arc<Mutex<HashMap<u32, String>>>,
collected_piece_tx: Sender<CollectedPiece>,
collected_piece_timeout: Duration,
) -> Result<CollectedParent> {
@ -481,33 +441,18 @@ impl PersistentCachePieceCollector {
);
})? {
let message = message?;
if let Some(mut parents) = collected_pieces.get_mut(&message.number) {
parents.push(parent.clone());
} else {
continue;
}
// Wait for collecting the piece from different parents when the first
// piece is collected.
tokio::time::sleep(DEFAULT_WAIT_FOR_PIECE_FROM_DIFFERENT_PARENTS).await;
let parents = match collected_pieces.remove(&message.number) {
Some((_, parents)) => parents,
None => continue,
};
let parent = match parents.get(fastrand::usize(..parents.len())) {
Some(parent) => parent,
None => {
error!(
"collected_pieces does not contain parent for piece {}",
message.number
);
// Remove the piece from collected_pieces, avoid to collect the same piece from
// different parents.
{
let mut collected_pieces_guard = collected_pieces.lock().await;
if collected_pieces_guard.remove(&message.number).is_none() {
continue;
}
};
}
info!(
"picked up piece {}-{} metadata from parent {}",
"received persistent cache piece {}-{} metadata from parent {}",
task_id, message.number, parent.id
);
@ -548,7 +493,11 @@ impl PersistentCachePieceCollector {
info!("peer {} sync persistent cache pieces finished", peer.id);
// If all pieces are collected, abort all tasks.
if collected_pieces.is_empty() {
let collected_pieces_guard = collected_pieces.lock().await;
let is_empty = collected_pieces_guard.is_empty();
drop(collected_pieces_guard);
if is_empty {
info!("all persistent cache pieces are collected, abort all tasks");
join_set.abort_all();
}

View File

@ -20,8 +20,7 @@ use crate::metrics::{
collect_backend_request_started_metrics,
};
use dragonfly_api::common::v2::{
Download, Hdfs, ObjectStorage, Peer, Piece, SizeScope, Task as CommonTask, TaskType,
TrafficType,
Download, Hdfs, ObjectStorage, Peer, Piece, Task as CommonTask, TrafficType,
};
use dragonfly_api::dfdaemon::{
self,
@ -49,7 +48,6 @@ use dragonfly_client_util::{
id_generator::IDGenerator,
};
use reqwest::header::HeaderMap;
use std::collections::HashMap;
use std::path::Path;
use std::sync::{
atomic::{AtomicBool, Ordering},
@ -241,14 +239,20 @@ impl Task {
// store the task.
if !task.is_finished() && !self.storage.has_enough_space(content_length)? {
return Err(Error::NoSpace(format!(
"not enough space to store the task: content_length={}",
"not enough space to store the persistent cache task: content_length={}",
content_length
)));
}
let task = self
.storage
.download_task_started(id, piece_length, content_length, response.http_header)
.download_task_started(
id,
piece_length,
content_length,
response.http_header,
request.load_to_cache,
)
.await;
// Attempt to create a hard link from the task file to the output path.
@ -388,7 +392,6 @@ impl Task {
range: request.range,
response_header: task.response_header.clone(),
pieces,
is_finished: task.is_finished(),
},
),
),
@ -734,6 +737,7 @@ impl Task {
remaining_interested_pieces.clone(),
request.is_prefetch,
request.need_piece_content,
request.load_to_cache,
download_progress_tx.clone(),
in_stream_tx.clone(),
)
@ -977,6 +981,7 @@ impl Task {
interested_pieces: Vec<metadata::Piece>,
is_prefetch: bool,
need_piece_content: bool,
load_to_cache: bool,
download_progress_tx: Sender<Result<DownloadTaskResponse, Status>>,
in_stream_tx: Sender<AnnouncePeerRequest>,
) -> ClientResult<Vec<metadata::Piece>> {
@ -1037,6 +1042,7 @@ impl Task {
finished_pieces: Arc<Mutex<Vec<metadata::Piece>>>,
is_prefetch: bool,
need_piece_content: bool,
load_to_cache: bool,
) -> ClientResult<metadata::Piece> {
// Limit the concurrent piece count.
let _permit = semaphore.acquire().await.unwrap();
@ -1057,6 +1063,7 @@ impl Task {
length,
parent.clone(),
is_prefetch,
load_to_cache,
)
.await
.map_err(|err| {
@ -1190,6 +1197,7 @@ impl Task {
finished_pieces.clone(),
is_prefetch,
need_piece_content,
load_to_cache,
)
.in_current_span(),
);
@ -1303,6 +1311,7 @@ impl Task {
request_header: HeaderMap,
is_prefetch: bool,
need_piece_content: bool,
load_to_cache: bool,
piece_manager: Arc<piece::Piece>,
semaphore: Arc<Semaphore>,
download_progress_tx: Sender<Result<DownloadTaskResponse, Status>>,
@ -1326,6 +1335,7 @@ impl Task {
length,
request_header,
is_prefetch,
load_to_cache,
object_storage,
hdfs,
)
@ -1430,6 +1440,7 @@ impl Task {
request_header.clone(),
request.is_prefetch,
request.need_piece_content,
request.load_to_cache,
self.piece.clone(),
semaphore.clone(),
download_progress_tx.clone(),
@ -1577,11 +1588,6 @@ impl Task {
}
};
if !piece.is_finished() {
debug!("piece {} is not finished, skip it", piece_id);
continue;
}
// Fake the download from the local.
self.piece.download_from_local(task_id, piece.length);
info!("finished piece {} from local", piece_id,);
@ -1696,6 +1702,7 @@ impl Task {
length: u64,
request_header: HeaderMap,
is_prefetch: bool,
load_to_cache: bool,
piece_manager: Arc<piece::Piece>,
semaphore: Arc<Semaphore>,
download_progress_tx: Sender<Result<DownloadTaskResponse, Status>>,
@ -1718,6 +1725,7 @@ impl Task {
length,
request_header,
is_prefetch,
load_to_cache,
object_storage,
hdfs,
)
@ -1776,6 +1784,7 @@ impl Task {
interested_piece.length,
request_header.clone(),
request.is_prefetch,
request.load_to_cache,
self.piece.clone(),
semaphore.clone(),
download_progress_tx.clone(),
@ -1821,74 +1830,7 @@ impl Task {
/// stat_task returns the task metadata.
#[instrument(skip_all)]
pub async fn stat(
&self,
task_id: &str,
host_id: &str,
local_only: bool,
) -> ClientResult<CommonTask> {
if local_only {
let Some(task_metadata) = self.storage.get_task(task_id).inspect_err(|err| {
error!("get task {} from local storage error: {:?}", task_id, err);
})?
else {
return Err(Error::TaskNotFound(task_id.to_owned()));
};
let piece_metadatas = self.piece.get_all(task_id).inspect_err(|err| {
error!(
"get pieces for task {} from local storage error: {:?}",
task_id, err
);
})?;
let pieces = piece_metadatas
.into_iter()
.filter(|piece| piece.is_finished())
.map(|piece| {
// The traffic_type indicates whether the first download was from the source or hit the remote peer cache.
// If the parent_id exists, the piece was downloaded from a remote peer. Otherwise, it was
// downloaded from the source.
let traffic_type = match piece.parent_id {
None => TrafficType::BackToSource,
Some(_) => TrafficType::RemotePeer,
};
Piece {
number: piece.number,
parent_id: piece.parent_id.clone(),
offset: piece.offset,
length: piece.length,
digest: piece.digest.clone(),
content: None,
traffic_type: Some(traffic_type as i32),
cost: piece.prost_cost(),
created_at: Some(prost_wkt_types::Timestamp::from(piece.created_at)),
}
})
.collect::<Vec<Piece>>();
return Ok(CommonTask {
id: task_metadata.id,
r#type: TaskType::Standard as i32,
url: String::new(),
digest: None,
tag: None,
application: None,
filtered_query_params: Vec::new(),
request_header: HashMap::new(),
content_length: task_metadata.content_length.unwrap_or(0),
piece_count: pieces.len() as u32,
size_scope: SizeScope::Normal as i32,
pieces,
state: String::new(),
peer_count: 0,
has_available_peer: false,
created_at: Some(prost_wkt_types::Timestamp::from(task_metadata.created_at)),
updated_at: Some(prost_wkt_types::Timestamp::from(task_metadata.updated_at)),
});
}
pub async fn stat(&self, task_id: &str, host_id: &str) -> ClientResult<CommonTask> {
let task = self
.scheduler_client
.stat_task(StatTaskRequest {
@ -1969,7 +1911,7 @@ mod tests {
// Create a task and save it to storage.
let task_id = "test-task-id";
storage
.download_task_started(task_id, 1024, 4096, None)
.download_task_started(task_id, 1024, 4096, None, false)
.await
.unwrap();

View File

@ -16,14 +16,12 @@
use dragonfly_client_config::dfdaemon::Host;
use opentelemetry::{global, trace::TracerProvider, KeyValue};
use opentelemetry_otlp::{WithExportConfig, WithTonicConfig};
use opentelemetry_otlp::WithExportConfig;
use opentelemetry_sdk::{propagation::TraceContextPropagator, Resource};
use rolling_file::*;
use std::fs;
use std::path::PathBuf;
use std::str::FromStr;
use std::time::Duration;
use tonic::metadata::{MetadataKey, MetadataMap, MetadataValue};
use tracing::{info, Level};
use tracing_appender::non_blocking::WorkerGuard;
use tracing_opentelemetry::OpenTelemetryLayer;
@ -44,10 +42,7 @@ pub fn init_tracing(
log_dir: PathBuf,
log_level: Level,
log_max_files: usize,
otel_protocol: Option<String>,
otel_endpoint: Option<String>,
otel_path: Option<PathBuf>,
otel_headers: Option<reqwest::header::HeaderMap>,
jaeger_addr: Option<String>,
host: Option<Host>,
is_seed_peer: bool,
console: bool,
@ -102,69 +97,26 @@ pub fn init_tracing(
let env_filter = EnvFilter::try_from_default_env()
.unwrap_or_else(|_| EnvFilter::default().add_directive(log_level.into()));
// Enable console subscriber layer for tracing spawn tasks on `127.0.0.1:6669` when log level is TRACE.
let console_subscriber_layer = if log_level == Level::TRACE {
Some(console_subscriber::spawn())
} else {
None
};
let subscriber = Registry::default()
.with(env_filter)
.with(console_subscriber_layer)
.with(file_logging_layer)
.with(stdout_logging_layer);
// If OTLP protocol and endpoint are provided, set up OpenTelemetry tracing.
if let (Some(protocol), Some(endpoint)) = (otel_protocol, otel_endpoint) {
let otlp_exporter = match protocol.as_str() {
"grpc" => {
let mut metadata = MetadataMap::new();
if let Some(headers) = otel_headers {
for (key, value) in headers.iter() {
metadata.insert(
MetadataKey::from_str(key.as_str())
.expect("failed to create metadata key"),
MetadataValue::from_str(value.to_str().unwrap())
.expect("failed to create metadata value"),
);
}
}
let endpoint_url = url::Url::parse(&format!("http://{}", endpoint))
.expect("failed to parse OTLP endpoint URL");
opentelemetry_otlp::SpanExporter::builder()
.with_tonic()
.with_endpoint(endpoint_url)
.with_timeout(SPAN_EXPORTER_TIMEOUT)
.with_metadata(metadata)
.build()
.expect("failed to create OTLP exporter")
}
"http" | "https" => {
let mut endpoint_url = url::Url::parse(&format!("{}://{}", protocol, endpoint))
.expect("failed to parse OTLP endpoint URL");
if let Some(path) = otel_path {
endpoint_url = endpoint_url
.join(path.to_str().unwrap())
.expect("failed to join OTLP endpoint path");
}
opentelemetry_otlp::SpanExporter::builder()
.with_http()
.with_endpoint(endpoint_url.as_str())
.with_protocol(opentelemetry_otlp::Protocol::HttpJson)
.with_timeout(SPAN_EXPORTER_TIMEOUT)
.build()
.expect("failed to create OTLP exporter")
}
_ => {
panic!("unsupported OTLP protocol: {}", protocol);
}
// Setup jaeger layer.
if let Some(mut jaeger_addr) = jaeger_addr {
jaeger_addr = if jaeger_addr.starts_with("http://") {
jaeger_addr
} else {
format!("http://{}", jaeger_addr)
};
let otlp_exporter = opentelemetry_otlp::SpanExporter::builder()
.with_tonic()
.with_endpoint(jaeger_addr)
.with_timeout(SPAN_EXPORTER_TIMEOUT)
.build()
.expect("failed to create OTLP exporter");
let host = host.unwrap();
let provider = opentelemetry_sdk::trace::SdkTracerProvider::builder()
.with_batch_exporter(otlp_exporter)
@ -211,7 +163,6 @@ pub fn init_tracing(
subscriber.init();
}
std::panic::set_hook(Box::new(tracing_panic::panic_hook));
info!(
"tracing initialized directory: {}, level: {}",
log_dir.as_path().display(),