[
https://issues.apache.org/jira/browse/BEAM-14185?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17513819#comment-17513819
]
Thiago Nunes commented on BEAM-14185:
-------------------------------------
Without this fix, metadata tables would be dangling in the user's Cloud Spanner
Database. The dangling tables would count towards the tables per database limit
for Cloud Spanner, potentially preventing users from creating their own tables
in the existing database and prevent new jobs to be executed.
We propose that this fix is integrated in version 2.38.0 as a cherry pick.
We have tested this feature with our own verification framework for a long
enough time period and modified integration tests to account for the
modification.
> [SpannerIO.readChangeStreams] Drop metadata tables at the end of the job
> ------------------------------------------------------------------------
>
> Key: BEAM-14185
> URL: https://issues.apache.org/jira/browse/BEAM-14185
> Project: Beam
> Issue Type: New Feature
> Components: io-java-gcp
> Affects Versions: 2.37.0
> Reporter: Thiago Nunes
> Assignee: Thiago Nunes
> Priority: P2
> Time Spent: 1h 20m
> Remaining Estimate: 0h
>
> The SpannerIO.readChangeStreams Connector uses metadata tables to keep track
> of its internal state during execution. At the moment, these metadata tables
> linger after the execution, meaning that users will have to drop them
> manually.
> In this change, we would like to drop them automatically once the job
> finishes. This should only occur after all partitions have been processed
> successfully and marked as finished.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
