[
https://issues.apache.org/jira/browse/CASSANDRA-4730?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13467167#comment-13467167
]
Arya Goudarzi commented on CASSANDRA-4730:
------------------------------------------
>1. there is no action you can take to mitigate given that the commitlog entry
>is probably not readable in the first place
I spent 2 days trying to figure out why I had a data loss during shutdowns
until I was told about this. I compiled a version with logging, and tried to
reproduce the scenario which resulted in missing data, and I was able to at
least get a notification of what happened that caused commit log not to get
replayed. If I had that hint earlier, something could have triggered in my head
earlier to try and tune commit log sync settings better so I don't get data
loss.
>2. checksum mismatch is not even a reliable indicator of unclean shutdown
Please advice, what can be a reliable indicator as I really think this is a
critical scenario to detect.
> CommitLogReplayer should report the bad CRC checksum in the log
> ---------------------------------------------------------------
>
> Key: CASSANDRA-4730
> URL: https://issues.apache.org/jira/browse/CASSANDRA-4730
> Project: Cassandra
> Issue Type: Bug
> Affects Versions: 1.1.5
> Environment: Cassandra 1.1.5
> Reporter: Arya Goudarzi
>
> If commit log isn't fully fsynced, the record which fails the checksum is not
> replayed do this logic. It would be beneficial to log that as an error so
> that user can know easily what happened.
> "./src/java/org/apache/cassandra/db/commitlog/CommitLogReplayer.java" line
> 188 of 287 --65%-- col 54
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
