[
https://issues.apache.org/jira/browse/IMPALA-14073?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18007330#comment-18007330
]
Michael Smith commented on IMPALA-14073:
----------------------------------------
One simpler option might be to retry "Table already exists" or "Table does not
exist" failures. Fixing the behavior itself is likely quite a bit harder, and
probably a rare occurence.
> test_local_catalog_ddls_with_invalidate_metadata_sync_ddl is flaky after
> IMPALA-13631
> -------------------------------------------------------------------------------------
>
> Key: IMPALA-14073
> URL: https://issues.apache.org/jira/browse/IMPALA-14073
> Project: IMPALA
> Issue Type: Bug
> Components: Catalog, Test
> Affects Versions: Impala 5.0.0
> Reporter: Riza Suminto
> Priority: Major
> Labels: broken-build
>
> IMPALA-13631 change alterTableOrViewRename to not hold catalog version lock
> while making HMS RPC.
> It adds back-and-forth ALTER TABLE RENAME testcase into
> test_local_catalog_ddls_with_invalidate_metadata_sync_ddl, which make it
> flaky for causing DDL occasionally not finding the target table after
> concurrent RENAME-INVALIDATE-REFRESH.
> We need to understand what happen and how to make the test deterministic.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
