cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Joel Knighton (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-11811) dtest failure in snapshot_test.TestArchiveCommitlog.test_archive_commitlog
Date Thu, 22 Sep 2016 18:07:20 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-11811?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15514027#comment-15514027
] 

Joel Knighton commented on CASSANDRA-11811:
-------------------------------------------

Both these failures occur because a commitlog from another node is replayed that contains
a mutation inserting tokens to the system.local table. This is why we see an exactly double
token count as configured in {{cassandra.yaml}}; the node generates its own tokens during
startup, and it inserts another collection of tokens during commitlog replay.

In the case of the CDC test, this is because the commitlog segment for this insert also contains
writes to the CDC table. We can fix this by draining the node and stopping/starting it before
doing any inserts to the CDC table.

In the case of the commitlog test, we remove archived commitlogs before snapshotting, but
we haven't ensured that the commitlog containing these system writes has been archived. Again,
one way we could achieve this by draining/stopping/starting the node before removing archived
commitlogs.

> dtest failure in snapshot_test.TestArchiveCommitlog.test_archive_commitlog
> --------------------------------------------------------------------------
>
>                 Key: CASSANDRA-11811
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-11811
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: Philip Thompson
>            Assignee: DS Test Eng
>              Labels: dtest
>             Fix For: 3.x
>
>
> example failure:
> http://cassci.datastax.com/job/trunk_dtest_win32/416/testReport/snapshot_test/TestArchiveCommitlog/test_archive_commitlog
> Failed on CassCI build trunk_dtest_win32 #416
> Relevant error is pasted. This is clearly a test problem. No idea why it only happens
on windows, as of yet. Affecting most tests in the TestArchiveCommitlog suite
> {code}
> WARN: Failed to flush node: node1 on shutdown.
> Unexpected error in node1 log, error: 
> ERROR [main] 2016-05-13 21:15:02,701 CassandraDaemon.java:729 - Fatal configuration error
> org.apache.cassandra.exceptions.ConfigurationException: Cannot change the number of tokens
from 64 to 32
> 	at org.apache.cassandra.service.StorageService.joinTokenRing(StorageService.java:1043)
~[main/:na]
> 	at org.apache.cassandra.service.StorageService.initServer(StorageService.java:740) ~[main/:na]
> 	at org.apache.cassandra.service.StorageService.initServer(StorageService.java:625) ~[main/:na]
> 	at org.apache.cassandra.service.CassandraDaemon.setup(CassandraDaemon.java:368) [main/:na]
> 	at org.apache.cassandra.service.CassandraDaemon.activate(CassandraDaemon.java:583) [main/:na]
> 	at org.apache.cassandra.service.CassandraDaemon.main(CassandraDaemon.java:712) [main/:na]
> {code}



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message