Alexey Serbin created KUDU-3424:
-----------------------------------
Summary: IsAlterTableDone RPC isn't working properly for
multi-replica tables when setting extra tablet configs
Key: KUDU-3424
URL: https://issues.apache.org/jira/browse/KUDU-3424
Project: Kudu
Issue Type: Bug
Affects Versions: 1.16.0
Reporter: Alexey Serbin
I noticed flakiness in {{TabletHistoryGcITest.TestSnapshotScanBeforeAHM}},
where failures manifested themselves with the following error:
{noformat}
src/kudu/integration-tests/tablet_history_gc-itest.cc:172
Failed
Bad status: Invalid argument: snapshot scan end timestamp is earlier than the
ancient history mark. Consider increasing the value of the configuration
parameter --tablet_history_max_age_sec. Snapshot timestamp: P: 1669591267778185
usec, L: 0 Ancient History Mark: P: 1669591267778244 usec, L: 0 Physical time
difference: -0.000s
{noformat}
It turned out the root case of the flakiness was not-yet-updated extra
configuration for a tablet when scan request at snapshot arrived.
Setting the number of replicas to one for the test table addressed the issue.
It's necessary to fix the issue with {{IsAlterTableDone}}. I guess there is a
bug affecting not only setting the extra configurations for a table, but other
DDL-related properties as well.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
