Closed mneverov closed 9 months ago
Thank you for the PR, much appreciated
The first problem is clearly a regression caused by this change. The only thing I would change is putting wait_backup
after verify_archive_mode_is_on
(as it used to be before the mentioned PR).
As for the second issue, I see the explanation slightly different.
Clone target is actually set to +1 minute. Problems appear when the tests are slow enough to include creation and then not to include drop of the table_with_oids
into the archive we use for the clone. Here I would rather try to avoid 'table with oids' making into clones at all. So maybe just delete it right after test_pg_upgrade_to_13_check_failed
? And probably also execute wait_backup
and wait_zero_lag
before the clone which were also moved to the wrong place in the mentioned PR
Thank you for the PR, much appreciated
The first problem is clearly a regression caused by this change. The only thing I would change is putting
wait_backup
afterverify_archive_mode_is_on
(as it used to be before the mentioned PR).As for the second issue, I see the explanation slightly different. Clone target is actually set to +1 minute. Problems appear when the tests are slow enough to include creation and then not to include drop of the
table_with_oids
into the archive we use for the clone. Here I would rather try to avoid 'table with oids' making into clones at all. So maybe just delete it right aftertest_pg_upgrade_to_13_check_failed
? And probably also executewait_backup
andwait_zero_lag
before the clone which were also moved to the wrong place in the mentioned PR
Hi @hughcapet, thank you for the thorough review. I applied your suggestions, ptal.
Fix flaky tests
Previously, when running on GH test failed periodically.
1 minute in the past
, so if tests were quick enough thedrop table with oids
statement is absent in backup yet and hence the table is cloned intoupgrade_container
. Which leads to the:Introducing waiting a backup fixed the first error, and dropping table with OIDs in
upgrade_container
fixed the second.