[
https://issues.apache.org/jira/browse/KUDU-2760?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Adar Dembo updated KUDU-2760:
-----------------------------
Affects Version/s: 1.10.0
Component/s: test
When this fails, it appears to fail on key 151 or 201, so perhaps there's a
flush boundary interplay?
Also, it may also be an artifact of the scans in the test not picking up the
newly written data. AFAICT the scans use {{READ_LATEST}} with
{{CLOSEST_REPLICA}}, so it's possible for the scan to land on a follower that's
a bit behind the majority and miss some data.
[~acelyc111] since you added these tests not long ago, could you take a look?
> kudu-tool-test.TestCopyTable is flaky
> -------------------------------------
>
> Key: KUDU-2760
> URL: https://issues.apache.org/jira/browse/KUDU-2760
> Project: Kudu
> Issue Type: Bug
> Components: test
> Affects Versions: 1.10.0
> Reporter: Hao Hao
> Priority: Major
> Attachments: kudu-tool-test.0.txt
>
>
> Encountered a failure of TestCopyTable in kudu-tool-test.cc with the
> following error:
> {noformat}
> I0403 04:31:25.699993 7865 catalog_manager.cc:3977] T
> 216c0526bd944c7da8c2a62cabe430ba P c80b8c54e90346559bbb413ebdb7d08f reported
> cstate change: term changed from 0 to 1, leader changed from <none> to
> c80b8c54e90346559bbb413ebdb7d08f (127.3.136.65). New cstate: current_term: 1
> leader_uuid: "c80b8c54e90346559bbb413ebdb7d08f" committed_config {
> opid_index: -1 OBSOLETE_local: true peers { permanent_uuid:
> "c80b8c54e90346559bbb413ebdb7d08f" member_type: VOTER last_known_addr { host:
> "127.3.136.65" port: 35451 } health_report { overall_health: HEALTHY } } }
> /data/somelongdirectorytoavoidrpathissues/src/kudu/src/kudu/tools/kudu-tool-test.cc:582:
> Failure
> Value of: dst_line
> Expected: has no substring "key"
> Actual: "(int32 key=151, int32 int_val=-325474784, string
> string_val=\"7ca8cde3dcca640a\")" (of type std::string)
> /data/somelongdirectorytoavoidrpathissues/src/kudu/src/kudu/tools/kudu-tool-test.cc:3067:
> Failure
> Expected: RunCopyTableCheck(arg) doesn't generate new fatal failures in the
> current thread.
> Actual: it does.
> {noformat}
> Attached the full log.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
