cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jonathan Ellis (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (CASSANDRA-6496) Endless L0 LCS compactions
Date Tue, 17 Dec 2013 14:40:08 GMT

     [ https://issues.apache.org/jira/browse/CASSANDRA-6496?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Jonathan Ellis updated CASSANDRA-6496:
--------------------------------------

    Attachment: 6496.txt

Patch to remove sstable output size limit when we're supposed to be doing STCS in L0.

- Chose to use LCT w/ unlimited size instead of normal CT since that seems less fragile (e.g.
if we decide CT.level() should return -1)
- Some churn to standardize on limiting in Bytes over MB


> Endless L0 LCS compactions
> --------------------------
>
>                 Key: CASSANDRA-6496
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-6496
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Core
>         Environment: Cassandra 2.0.3, Linux, 6 nodes, 5 disks per node
>            Reporter: Nikolai Grigoriev
>         Attachments: 6496.txt, system.log.1.gz, system.log.gz
>
>
> I have first described the problem here: http://stackoverflow.com/questions/20589324/cassandra-2-0-3-endless-compactions-with-no-traffic
> I think I have really abused my system with the traffic (mix of reads, heavy updates
and some deletes). Now after stopping the traffic I see the compactions that are going on
endlessly for over 4 days.
> For a specific CF I have about 4700 sstable data files right now.  The compaction estimates
are logged as "[3312, 4, 0, 0, 0, 0, 0, 0, 0]". sstable_size_in_mb=256.  3214 files are about
256Mb (+/1 few megs), other files are smaller or much smaller than that. No sstables are larger
than 256Mb. What I observe is that LCS picks 32 sstables from L0 and compacts them into 32
sstables of approximately the same size. So, what my system is doing for last 4 days (no traffic
at all) is compacting groups of 32 sstables into groups of 32 sstables without any changes.
Seems like a bug to me regardless of what did I do to get the system into this state...



--
This message was sent by Atlassian JIRA
(v6.1.4#6159)

Mime
View raw message