[
https://issues.apache.org/jira/browse/KUDU-3813?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18124231#comment-18124231
]
ASF subversion and git services commented on KUDU-3813:
-------------------------------------------------------
Commit 6a51400c44c7d0966fcf01950bde2d5776fbbce6 in kudu's branch
refs/heads/master from Alexey Serbin
[ https://gitbox.apache.org/repos/asf?p=kudu.git;h=6a51400c4 ]
KUDU-3813 fix flakiness in txn_commit-itest scenarios
With very short keepalive interval setting for transactions, test
scenarios would fail from time to time in TSAN builds. For example,the
TestDontBackgroundAbortIfCommitInProgress scenario would intermittently
fail as described in the JIRA item. This changelist increases the
keepalive interval from 300ms to 500ms, and also increases verification
time interval from 100ms to 250ms. With this, a wider margin is
provided to cover for scheduler anomalies and other TSAN-induced pauses
at overloaded test nodes.
I ran 5K iterations of the TestDontBackgroundAbortIfCommitInProgress
scenario and about 500 iterations of all the scenarios in TSAN
instrumented txn_commit-itest with --stress-cpu-threads=8 after this
update and there were no failures reported.
Change-Id: I58746e38866788c0ec6b9ebac96bc8ec1c23ca72
Reviewed-on: http://gerrit.cloudera.org:8080/25029
Tested-by: Alexey Serbin <[email protected]>
Reviewed-by: Alexey Serbin <[email protected]>
> TxnCommitITest.TestDontBackgroundAbortIfCommitInProgress fails from time to
> time
> --------------------------------------------------------------------------------
>
> Key: KUDU-3813
> URL: https://issues.apache.org/jira/browse/KUDU-3813
> Project: Kudu
> Issue Type: Bug
> Components: test, transactions
> Affects Versions: 1.19.0
> Reporter: Alexey Serbin
> Priority: Major
> Attachments: txn_commit-itest.txt.xz
>
>
> The TestDontBackgroundAbortIfCommitInProgress scenario in txn_commit-itest
> fails from time to time with errors like below (at least in TSAN builds):
> {noformat}
> I20261005 12:23:20.561852 20844 txn_status_manager.cc:1391] automatically
> aborted stale txn (ID 0) past 0.305s from last keepalive heartbeat (effective
> timeout is 0.300s)
> src/kudu/integration-tests/txn_commit-itest.cc:841: Failure
> Failed
> Bad status: Illegal state: Failed to write to server:
> eac03b2130bb42d2bdd422415fd97dfc (127.16.33.65:38953): transaction ID 0 is
> not open: state: ABORT_IN_PROGRESS user: "slave" start_timestamp: 1791203000
> last_transition_timestamp: 1791203000
> {noformat}
> It seems the problem is that the started test transaction is automatically
> aborted because of staleness -- perhaps, scheduler anomalies and/or delays of
> overloaded test node is the root cause of the flakiness. Increasing
> keepalive timeout should help.
> Full log is attached.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
