cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Benedict (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-9669) If sstable flushes complete out of order, on restart we can fail to replay necessary commit log records
Date Fri, 13 May 2016 23:47:13 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-9669?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15283315#comment-15283315
] 

Benedict commented on CASSANDRA-9669:
-------------------------------------

I'm not sure if the behaviour for locking on the base table when flushing indexes was present
when I wrote this patch, but both that sync (in SecondaryIndexManager), and this one (in truncation),
are ill advised:

* Synchronizing on a base table for an action that is triggered by an acton that itself must
synchronize on the base table is asking for trouble
* Synchronizing in either case over a lengthy operation that may itself synchronize is also
asking for trouble

As it is, I threw the synchronized in there along with some other modifications to make the
truncation of views less obviously broken.  I think it's probably still broken, just less
obviously so, so it may as well be taken out to make this right.

> If sstable flushes complete out of order, on restart we can fail to replay necessary
commit log records
> -------------------------------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-9669
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-9669
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Local Write-Read Paths
>            Reporter: Benedict
>            Assignee: Benedict
>            Priority: Critical
>              Labels: correctness
>             Fix For: 2.2.7, 3.7, 3.0.7
>
>
> While {{postFlushExecutor}} ensures it never expires CL entries out-of-order, on restart
we simply take the maximum replay position of any sstable on disk, and ignore anything prior.

> It is quite possible for there to be two flushes triggered for a given table, and for
the second to finish first by virtue of containing a much smaller quantity of live data (or
perhaps the disk is just under less pressure). If we crash before the first sstable has been
written, then on restart the data it would have represented will disappear, since we will
not replay the CL records.
> This looks to be a bug present since time immemorial, and also seems pretty serious.



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

Mime
View raw message