c37ea15830
core: also expose parser on JsonParser
2018-04-20 13:25:21 -07:00
74532acce0
core: use List instead of Set for drainedSubstreams
...
This improves latency performance (retry enabled) for NETTY transport by about 2us. For INPROCESS, NETTY_LOCAL, OKHTTP transports, the improvement seems much less than 2us.
```
before
Benchmark (direct) (transport) Mode Cnt Score Error Units
TransportBenchmark.unaryCall1024 true NETTY sample 260801 76566.536 ± 189.439 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.00 true NETTY sample 60480.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.50 true NETTY sample 75264.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.90 true NETTY sample 85504.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.95 true NETTY sample 87424.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.99 true NETTY sample 100864.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.999 true NETTY sample 140800.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.9999 true NETTY sample 205547.469 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p1.00 true NETTY sample 3915776.000 ns/op
TransportBenchmark.unaryCall1024 false NETTY sample 208352 95865.619 ± 113.142 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.00 false NETTY sample 72576.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.50 false NETTY sample 93568.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.90 false NETTY sample 105728.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.95 false NETTY sample 109568.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.99 false NETTY sample 124544.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.999 false NETTY sample 161792.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.9999 false NETTY sample 230520.448 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p1.00 false NETTY sample 3997696.000 ns/op
```
```
after
Benchmark (direct) (transport) Mode Cnt Score Error Units
TransportBenchmark.unaryCall1024 true NETTY sample 269471 74104.666 ± 182.514 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.00 true NETTY sample 60992.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.50 true NETTY sample 70912.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.90 true NETTY sample 83584.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.95 true NETTY sample 85888.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.99 true NETTY sample 100224.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.999 true NETTY sample 142848.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.9999 true NETTY sample 489527.706 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p1.00 true NETTY sample 4300800.000 ns/op
TransportBenchmark.unaryCall1024 false NETTY sample 216000 92468.337 ± 90.229 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.00 false NETTY sample 68480.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.50 false NETTY sample 89472.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.90 false NETTY sample 103680.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.95 false NETTY sample 107008.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.99 false NETTY sample 120960.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.999 false NETTY sample 159232.000 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p0.9999 false NETTY sample 272076.493 ns/op
TransportBenchmark.unaryCall1024:unaryCall1024·p1.00 false NETTY sample 2662400.000 ns/op
```
2018-04-20 11:22:32 -07:00
ddf77f59cc
core: Fix experimental API issue for channel state API
...
The earlier issue was for the feature itself, not the stabilization.
2018-04-19 17:38:20 -07:00
934c2e2fd7
core: clarify binlog API ( #4362 )
...
The interceptors are a part of the implementation and not meant for
callers.
Both methods for wrapping clients / servers can be `final`.
2018-04-19 17:32:45 -07:00
ec95999124
core: remove unneeded binlog tests for ManagedChannelImpl ServerImpl ( #4361 )
...
Replace the tests with tests that just make sure binary log is
installed. Better tests exist today in BinaryLogProviderTest to make
sure the actual logic is correct.
2018-04-19 16:55:32 -07:00
e67a61fa08
core: refactor ManagedChannelImplTest with a configurable builder
...
In previous code of `ManagedChannelImplTest`, new tests will have to add more and more custom args in `createChannel(...)`, and each time when a custom arg is needed all other default args need be repeatedly provided in `createChannel(...)`. In most cases those args are just builder attributes. So this PR makes ManagedChannelImplTest easier to stub by providing a builder field, then new tests will just set necessary builder attributes before calling `createChannel()`.
2018-04-19 10:06:39 -07:00
f6a032d6b2
core: forward toString for forwarding classes
...
Forward `toString()` method for forwarding classes to improve debug information.
For example, `ForwardingManagedChannel.toString()` will return something like:
```
ForwardingManagedChannel{delegate=ManagedChannelImpl{logId=tag-13247, target=localhost:8080}}
```
2018-04-19 10:05:48 -07:00
f754bff906
core: make JsonParser public
2018-04-18 18:49:08 -07:00
30400e0e58
core: add sockets to subchannel stat ( #4345 )
...
This is info not being reported at the moment and is missing from the
display.
2018-04-16 11:33:28 -07:00
c04278ea07
services: turn channelz null futures into Status.UNIMPLEMENTED ( #4346 )
...
Then channelz GUI will take this into account. This is particularly
useful for InProcessTransport, where I have decided we do not need
special support for in channelz. The server and channel stats are
already sufficient.
2018-04-16 09:54:35 -07:00
791a29f63b
services: socket options were erroneously ignored if socket has no stats ( #4336 )
...
This fixes listen sockets. It is ok to have no data but report socket
options.
2018-04-13 14:59:35 -07:00
e4502aca5d
stub: add docs clarifying readiness
2018-04-13 14:12:54 -07:00
03465a7f45
core: don't warn on absent service config
...
Fixes #4333
2018-04-13 11:13:57 -07:00
a6811d5bb1
all: add toString() to Server to improve debug info
2018-04-13 09:41:09 -07:00
0eafb8e9ca
core: compare doubles with compare instead of ==
...
Always not compare equality for doubles with `==`.
2018-04-13 09:39:59 -07:00
0a598c053c
core: Fix lint, style, "good ideas" violations
2018-04-12 17:26:54 -07:00
6d7e19cbe5
core: replumb RetryPolicy
...
- replumbed `RetryPolicy` with `MethodInfo` without breaking the existing `RetryPolicyTest`.
- moved `ServiceConfigInterceptor.MethodInfo.RetryPolicy` out as a top level class so that `RetriableStream` does not import `ServiceConfigInterceptor`.
2018-04-12 11:41:04 -07:00
3f63cd0ad4
core: always call StreamTracer.streamClosed() when stream is officially closed. ( #4331 )
...
Previously StreamTracer.streamClosed() is called in
ServerStream.close(), but it is not exactly when the stream is
officially closed. ServerStreamListener.closed() is guaranteed to be
called and it is the official end of the stream.
2018-04-12 09:05:48 -07:00
137e759fda
core, netty: allow InputStream based certs ( #4316 )
...
Allow ServerBuilder to read certs from InputStream, not just from a
File.
2018-04-11 16:14:58 -07:00
62cc2b1ae3
core,netty: add NettySocketSupport to populate TcpInfo ( #4306 )
...
NettySocketSupport is responsible for making the low level calls to
get and populate the TcpInfo structure.
2018-04-05 10:36:49 -07:00
2df76cc710
core: apply service config
2018-04-04 17:45:12 -07:00
9ed84258aa
core: don't reschedule idle timer if it is already active
...
Benchmark results (3 runs each)
non direct
```
Before:
50.0%ile Latency (in nanos): 157471
90.0%ile Latency (in nanos): 185927
95.0%ile Latency (in nanos): 195135
99.0%ile Latency (in nanos): 218815
99.9%ile Latency (in nanos): 1188735
100.0%ile Latency (in nanos): 18333695
QPS: 6126
50.0%ile Latency (in nanos): 160407
90.0%ile Latency (in nanos): 188551
95.0%ile Latency (in nanos): 197487
99.0%ile Latency (in nanos): 219575
99.9%ile Latency (in nanos): 390239
100.0%ile Latency (in nanos): 18338815
QPS: 6106
50.0%ile Latency (in nanos): 157831
90.0%ile Latency (in nanos): 186439
95.0%ile Latency (in nanos): 195815
99.0%ile Latency (in nanos): 216951
99.9%ile Latency (in nanos): 281167
100.0%ile Latency (in nanos): 5384447
QPS: 6235
After:
50.0%ile Latency (in nanos): 152255
90.0%ile Latency (in nanos): 180551
95.0%ile Latency (in nanos): 188943
99.0%ile Latency (in nanos): 209623
99.9%ile Latency (in nanos): 1184831
100.0%ile Latency (in nanos): 4351999
QPS: 6313
50.0%ile Latency (in nanos): 153663
90.0%ile Latency (in nanos): 181671
95.0%ile Latency (in nanos): 189991
99.0%ile Latency (in nanos): 210495
99.9%ile Latency (in nanos): 278895
100.0%ile Latency (in nanos): 18283519
QPS: 6300
50.0%ile Latency (in nanos): 152767
90.0%ile Latency (in nanos): 180839
95.0%ile Latency (in nanos): 189791
99.0%ile Latency (in nanos): 211719
99.9%ile Latency (in nanos): 280927
100.0%ile Latency (in nanos): 12231167
QPS: 6381
```
direct:
```
Before:
50.0%ile Latency (in nanos): 133943
90.0%ile Latency (in nanos): 153671
95.0%ile Latency (in nanos): 163655
99.0%ile Latency (in nanos): 188871
99.9%ile Latency (in nanos): 235791
100.0%ile Latency (in nanos): 7864575
QPS: 7134
50.0%ile Latency (in nanos): 131623
90.0%ile Latency (in nanos): 151863
95.0%ile Latency (in nanos): 162095
99.0%ile Latency (in nanos): 187719
99.9%ile Latency (in nanos): 234983
100.0%ile Latency (in nanos): 17836031
QPS: 7250
50.0%ile Latency (in nanos): 131223
90.0%ile Latency (in nanos): 150823
95.0%ile Latency (in nanos): 161311
99.0%ile Latency (in nanos): 187719
99.9%ile Latency (in nanos): 237471
100.0%ile Latency (in nanos): 4416255
QPS: 7273
After:
50.0%ile Latency (in nanos): 122751
90.0%ile Latency (in nanos): 140967
95.0%ile Latency (in nanos): 148911
99.0%ile Latency (in nanos): 173215
99.9%ile Latency (in nanos): 214823
100.0%ile Latency (in nanos): 18509823
QPS: 7774
50.0%ile Latency (in nanos): 124507
90.0%ile Latency (in nanos): 145855
95.0%ile Latency (in nanos): 156623
99.0%ile Latency (in nanos): 183111
99.9%ile Latency (in nanos): 235679
100.0%ile Latency (in nanos): 18289663
QPS: 7625
50.0%ile Latency (in nanos): 124295
90.0%ile Latency (in nanos): 145071
95.0%ile Latency (in nanos): 156439
99.0%ile Latency (in nanos): 183919
99.9%ile Latency (in nanos): 232447
100.0%ile Latency (in nanos): 3712383
QPS: 7632
```
2018-04-04 16:02:02 -07:00
8f01084bb3
core: add a `close` to InputBufferStream
...
Before:
`InputBufferStream.close()` does not close their buffer so the buffer will leak.
After:
Resolves #4198 .
Override the `close` for closing their buffer.
2018-04-03 16:11:19 -07:00
b1d91b9f60
core,services: binary log should use real peer socket and call id ( #4266 )
...
The peer socket is read from TRANSPORT_ATTR_REMOTE_ADDR from the
stream attributes. We only log the peer on receive initial metadata.
The call id assumes census is available. The call ID read from the
context via SERVER_CALL_ID_CONTEXT_KEY on server side, and read from
CallOptions via CLIENT_CALL_ID_CALLOPTION_KEY on client side. The
value is copied from CONTEXT_SPAN_KEY which is set by census.
Pass around CallId with two longs, not a byte[].
2018-04-03 13:39:11 -07:00
724e32fe57
core,netty,services: add server listen sockets to channelz proto service ( #4220 )
...
Server listen sockets differ from normal sockets in that they do not
have a remote address, do not have stats on calls started/failed/etc,
and do not have security info.
2018-04-03 10:05:47 -07:00
1d901a8de2
core: add tests for ReadableBuffers
2018-04-03 08:11:39 -07:00
0067415811
core: document how to set up squid proxy for testing ( #4285 )
...
This makes it easier for a gRPC developer to validate the proxy code
using a real proxy.
2018-03-30 11:28:10 -07:00
7c37f1d32b
core,netty,okhttp,services: expose socket options to channelz ( #4228 )
...
For okhttp, expose the standard options from the Socket object.
For netty, expose all the `io.netty.channel.ChannelOption`s of the
`channel.config()`.
2018-03-30 08:48:37 -07:00
4e82e62eaa
Fix compilation in Java 9
2018-03-28 17:13:39 -07:00
5af2515f3e
core,README: add docs about grpc-java-api-checker ( #4182 )
2018-03-28 15:34:18 -07:00
db64c36af2
core: no retry on dropped PickResult
...
Address #3553 for normal retry. Not covering the unimplemented hedging case.
2018-03-26 14:57:12 -07:00
5f0ee3a725
Bump to Guava 20
...
Fixes #4176
2018-03-23 15:45:05 -07:00
7eab0d9468
netty: Add support for Conscrypt
2018-03-23 10:05:45 -07:00
0955866d5a
core,services: implement binarylog Client and ServerInterceptor ( #4174 )
...
- To make unit tests easier, added the BinaryLogSinkWriter abstract
class, which allows verifying high level arguments rather than low
level protobufs.
- remove InputStream based marshaller (IDENTITY_MARSHALLER) in favor
of byte[] because retries will need to call serialize multiple
times. This is achieved with a special InputStream that lets us
access the internal byte[] without copying.
- Change ByteBuffer based internal methods to use byte[]
- Update unit tests to reflect the above
2018-03-23 09:29:24 -07:00
5337728fa7
core,netty: client sends rst stream when server half-closes ( #4222 )
2018-03-22 12:38:42 -07:00
7c69c3a867
core: disable jndi lookups
2018-03-21 12:55:54 -07:00
900b39635d
core: Fix java lint found during weekly import
2018-03-20 17:45:26 -07:00
1e0875dff7
core: re-enable panic mode after all affected internal tests have been fixed. ( #4245 )
2018-03-20 13:06:41 -07:00
6836b31fb3
core: catch and log Throwable for jndiResolver.resolve(host)
...
JndiResolver.resolve may throw Error that could cause rpc to hang. Catch and log Throwable and keep using jdkResolver's result to prevent it.
2018-03-20 13:02:52 -07:00
95917b8f28
core: parse retry policy from service config
2018-03-19 15:15:59 -07:00
bb0c0e53c2
core: perform channelz obj removal before things that may throw ( #4235 )
...
This way we won't risk an exception making us fail to clean up.
For transportClosed only remove if we know this isn't a
duplicate call to close.
2018-03-18 12:18:09 -07:00
ca99f80999
core,services: track server sockets in channelz ( #4226 )
...
Rather than querying the ServerImpl for its sockets, we register them
into channelz as with all the other entities, for consistency.
Error conditions are checked with `assert` statements, so we throw in
tests.
2018-03-16 17:21:25 -07:00
2ce1d370a0
core: re-add back in orphan wrapper
2018-03-16 14:51:38 -07:00
7daefd75a1
core: make ManagedChannel honor Service config LB
2018-03-16 14:47:25 -07:00
26fc6324c4
core: fix typo in ClientCall.java javadoc
2018-03-15 10:07:30 -07:00
2761bbb851
core: handle long dns txt records properly, parse service config, add tests
2018-03-13 17:37:04 -07:00
6b9e090d70
core: rm unnecessary InternalNameResolverProvider
2018-03-13 10:21:40 -07:00
11c8bcde1b
Start 1.12.0 development cycle
2018-03-13 09:04:25 -07:00
7852b34781
core: fix channel.getState(true) will not reconnect
...
resolves #4170
2018-03-12 14:13:52 -07:00
a83f67a706
core,netty,okhttp: Transparent retry
...
Changes:
- `ClientStreamListener.onClose(Status status, RpcProgress rpcProgress, Metadata trailers)` added.
- `AbstractClientStream.transportReportStatus(Status status, RpcProgress rpcProgress, boolean stopDelivery, Metadata trailers)` added
- `ClientCallImpl.ClientStreamListenerImpl` will ignore the arg `rpcProgress` (non retry)
- `RetriableStream.SubListener` will handle `rpcProgress` and decide if transparent retry.
- `NettyClientHandler` and `OkHttpClientTransport` will pass `RpcProgress.REFUSED` to client stream listener for later stream ids when received GOAWAY, or for stream received a RST_STREAM frame with REFUSED code.
- All other files are just a result of refactoring.
2018-03-12 14:12:46 -07:00
Carl Mastrangelo
ZHANG Dapeng
Eric Anderson
zpencer
Kun Zhang
Shohei Kamimori
Eric Gribkoff