cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "amorton (JIRA)" <>
Subject [jira] [Commented] (CASSANDRA-5469) Race condition between index building and scrubDirectories() at startup
Date Tue, 16 Apr 2013 21:19:16 GMT


amorton commented on CASSANDRA-5469:

I think this is a different problem. 

When the CFS's are created they will start rebuilding secondary indexes async. Flushing the
secondary indexes SSTbles is not done by the MeteredFlusher, see SecondaryIndex.buildIndexBlocking().
Then when scrubDirectories() runs it will delete those -tmp- files out from under the index
> Race condition between index building and scrubDirectories() at startup
> -----------------------------------------------------------------------
>                 Key: CASSANDRA-5469
>                 URL:
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Core
>    Affects Versions: 1.0.12, 1.1.10, 1.2.4
>            Reporter: amorton
> From user group
> In CassandraDaemon.setup() the call to SystemTable.checkHealth() results in the CFS's
being created. As part of their creation they kick of async secondary index build if the index
is not marked as built (SecondaryIndexManager.addIndexedColumn()). Later in CD.setup() the
call is made to scrub the data dirs and this can race with the tmp files created by the index
rebuild. The result is an error that prevents the node starting.
> Should we delay rebuilding secondary indexes until after startup has completed or rebuild
them synchronously ? 

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:

View raw message