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-7511) Commit log grows infinitely after truncate
Date Wed, 23 Jul 2014 22:40:39 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-7511?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14072479#comment-14072479
] 

Benedict commented on CASSANDRA-7511:
-------------------------------------

Good point. I've done this, and have another patch, however looking closely at the code I
think we have generally got truncation a bit wrong, and it is still subtly broken. Especially
with user-defined timestamps. We take the current server time in millis and drop any tables
with data that is strictly older than this. We could (and, I think, should) use the ReplayPosition
we store against each sstable. Unfortunately this might break custom 2is, but for those we
can stick to sys-time for now, since that's what they've been making do with, and we can fix
it there as well for 2.1

It's also not clear to me on reading this why we nuke the memtables without flushing, since
we seem intent on snapshotting the data (and in fact flush it twice). This bug affects the
case where we do not snapshot, however in the case that we do surely we want all mutations
in the memtables to hit disk prior to snapshotting, else the snapshot won't contain the records
(or, alternatively, we don't care about flushing _any_ of the data at all)

[~brandon.williams] wdyt?

> Commit log grows infinitely after truncate
> ------------------------------------------
>
>                 Key: CASSANDRA-7511
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-7511
>             Project: Cassandra
>          Issue Type: Bug
>         Environment: CentOS 6.5, Oracle Java 7u60, C* 2.0.6, 2.0.9, including earlier
1.0.* versions.
>            Reporter: Viktor Jevdokimov
>            Assignee: Benedict
>            Priority: Minor
>              Labels: commitlog
>             Fix For: 2.0.10
>
>         Attachments: 7511.txt
>
>
> Commit log grows infinitely after CF truncate operation via cassandra-cli, regardless
CF receives writes or not thereafter.
> CF's could be non-CQL Standard and Super column type. Creation of snapshots after truncate
is turned off.
> Commit log may start grow promptly, may start grow later, on a few only or on all nodes
at once.
> Nothing special in the system log. No idea how to reproduce.
> After rolling restart commit logs are cleared and back to normal. Just annoying to do
rolling restart after each truncate.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message