[
https://issues.apache.org/jira/browse/CASSANDRA-2427?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jonathan Ellis reassigned CASSANDRA-2427:
-----------------------------------------
Assignee: paul cannon
How about this:
- create a commitlog-space-cap setting; if space gets above that, flush every
dirty CF in the oldest segment and remove it
- remove the per-CF memtable_flush_after_minutes setting in favor of the above
(this fixes the flush-storm problem of all your CFs flushing at once when you
configure this too low)
- while we're at it, remove the commitlog segment size config setting, nobody
has ever changed it to my knowledge in 2+ years
(I don't think we want to add an additional check for "is there enough space
for a new segment" at segment creation time; if there isn't enough space for
that there probably isn't enough to flush either and we're screwed either way.
So it would be complexity w/o a purpose.)
> Heuristic or hard cap to prevent fragmented commit logs from bringing down
> the server
> -------------------------------------------------------------------------------------
>
> Key: CASSANDRA-2427
> URL: https://issues.apache.org/jira/browse/CASSANDRA-2427
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Benjamin Coverston
> Assignee: paul cannon
> Labels: commitlog, hardening
> Fix For: 1.0
>
>
> Widely divergent write rates on column families can cause the commit log
> segments to fragment. In some cases we have seen the commit log partition
> overrun.
> One solution here would be to create a heuristic for segment fragmentation to
> trigger a flush (commit log segments/memtable) or simply track the free disk
> space and force a global flush when the disk gets to 80% capacity.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
