[
https://issues.apache.org/jira/browse/BEAM-13373?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17452646#comment-17452646
]
Brian Hulette commented on BEAM-13373:
--------------------------------------
Ah, I think that is due to leaked databases, though. I don't think we're
running >100 tests concurrently.
> Python PostCommits produce too many spanner databases
> -----------------------------------------------------
>
> Key: BEAM-13373
> URL: https://issues.apache.org/jira/browse/BEAM-13373
> Project: Beam
> Issue Type: Bug
> Components: testing
> Reporter: Brian Hulette
> Assignee: Yichi Zhang
> Priority: P2
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> There are ~30 Python PostCommit tests that create and destroy a Spanner
> Database. Each running across 3 PostCommits. We should reduce the number of
> tests that do this.
> A couple of avenues:
> 1. Only run Spanner ITs in one Python version's PostCommit.
> 2. Verify more logic in spannerio_test.py (as opposed to
> {{spannerio_\*_it_test.py}}), which uses a mock rather than prod Spanner. We
> can still have a few ITs, but we don't need to verify *everything* there.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
