Artifact stager should validate the manifest at pipeline submission time.
- Dominant language
- Java
- Stars
- 8.7k
- Forks
- 4.7k
- Avg merge
- 1d 20h
- Merged PRs (30d)
- 196
Description
One step of executing a Beam pipeline is staging pipeline dependencies to the runner, for Python SDK example see: [https://github.com/apache/beam/blob/1e2218aebf4902ca2f9107c885ff7ef0e1ef6eb8/sdks/python/apache_beam/runners/portability/stager.py#L83](https://github.com/apache/beam/blob/1e2218aebf4902ca2f9107c885ff7ef0e1ef6eb8/sdks/python/apache_beam/runners/portability/stager.py#L83) and it's implementations.
As a part of this process, we create a manifest of all staged artifacts. We should identify the requirements that constitute correctness of the manifest (for example, artifacts do not repeat twice, see also: https://github.com/apache/beam/blob/b0b7e3bf4941f874d360923ffd1d03e38befc589/sdks/go/pkg/beam/artifact/gcsproxy/retrieval.go#L122), and verify these requirements on the SDK side at pipeline submission, to fail faster.
Imported from Jira [BEAM-4407](https://issues.apache.org/jira/browse/BEAM-4407). Original Jira may contain additional context.
Reported by: tvalentyn.
This issue has child subcomponents which were not migrated over. See the original Jira for more information.
Contributor guide
Assessment
This issue has not been assessed yet.