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 Tue, 15 Sep 2015 13:10:45 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-9669?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14745432#comment-14745432
] 

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

bq. Memtable.isCleanAfter doesn't look right. 

Good catch, I have inverted the implementation and meaning, and renamed it to {{mayContainDataSince}}

bq.  "ka" reader says compatible with "kb" data but read will fail

Hmm. This is a bit of a problem. I must admit I didn't look too closely at compatibility,
since I assumed the whole point of the major/minor chars was to permit intra- and extra-version
sstable version increments. Without that, this seems to be a bit of a mess. I guess we will
need to increment all of the sstable versions past the max of the current. We should perhaps
rethink accepting all versions <= first char of current, as it doesn't permit much flexibility.

Thanks. I'll address this, your nits, and what looks like a relatively innocuous problem with
DTCS shortly.





> 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: Core
>            Reporter: Benedict
>            Assignee: Benedict
>            Priority: Critical
>              Labels: correctness
>             Fix For: 3.x, 2.1.x, 2.2.x, 3.0.x
>
>
> 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