ada18bc6e6
Reduced the timeout from 30 seconds to 5. This should not be needed as most users tell us that pods work even when the cache service is not available. But there was at least one customer who experienced timeout failures when creating pods after the service was deleted, but not the webhook config. |
||
|---|---|---|
| .. | ||
| api | ||
| metadata_writer | ||
| src | ||
| test | ||
| Dockerfile | ||
| Dockerfile.bazel | ||
| Dockerfile.cacheserver | ||
| Dockerfile.persistenceagent | ||
| Dockerfile.scheduledworkflow | ||
| Dockerfile.viewercontroller | ||
| Dockerfile.visualization | ||
| OWNERS | ||
| README.md | ||
| build_api_server.sh | ||
| requirements.in | ||
| requirements.txt | ||
| update_requirements.sh | ||
README.md
This directory contains code for the components that comprise the Kubeflow Pipelines backend.
Bazel
The supported Bazel version is 0.24.0.
Building & Testing
All components can be built using Bazel. To build
everything under backend, run: bazel build --action_env=PATH --define=grpc_no_ares=true //backend/...
To run all tests: bazel test --action_env=PATH --define=grpc_no_ares=true //backend/...
The API server itself can only be built/tested using Bazel. The following
commands target building and testing just the API server. bazel build --action_env=PATH --define=grpc_no_ares=true backend/src/apiserver/... bazel test --action_env=PATH --define=grpc_no_ares=true backend/src/apiserver/...
Building APIServer Image using Remote Build Execution
If you are a dev in the Kubeflow Pipelines team, you can use Remote Build Execution Service to build the API Server image using Bazel with use of a shared cache for speeding up the build. To do so, execute the following command:
./build_api_server.sh -i gcr.io/cloud-ml-pipelines-test/api-server:dev
Building Go client library and swagger files
After making changes to proto files, the Go client libraries and swagger files need to be regenerated and checked-in. The backend/api/generate_api.sh script takes care of this.
Updating BUILD files
As the backend is written in Go, the BUILD files can be updated automatically
using Gazelle. Whenever a Go file
is added or updated, run the following to ensure the BUILD files are updated as
well: bazel run //:gazelle
If a new external Go dependency is added, or an existing one has its version
bumped in the go.mod file, ensure the BUILD files pick this up by updating the
WORKSPACE go_repository rules using the following command: bazel run //:gazelle -- update-repos --from_file=go.mod
Updating python dependencies
pip-tools is used to manage python
dependencies. To update dependencies, edit requirements.in
and run ./update_requirements.sh <requirements.in >requirements.txt to update and pin the transitive
dependencies.