[jira] [Commented] (CASSANDRA-11155) Scrub/Cleanup block snapshots

["Thomas Steinmaurer (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CASSANDRA-11155?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16047737#comment-16047737 ] 

Thomas Steinmaurer commented on CASSANDRA-11155:
------------------------------------------------

We are seeing the same issue with Cassandra 2.1.15. We have an hourly snapshot CRON job in place and recently extended our production cluster from 6 to 9 nodes, thus executing nodetool cleanup on the older 6 nodes one by one. What we see is:
* Accummulating CROND based nodetool snapshot processes according to the Linux process list
* Host CPU usage ~ 95% with a m4.xlarge in AWS, using 4 concurrent_compactors and the default 16 MB/s compaction threshold

> Scrub/Cleanup block snapshots
> -----------------------------
>
>                 Key: CASSANDRA-11155
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-11155
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Compaction
>         Environment: Linux Ubuntu hosts, Cassandra 2.1.12
>            Reporter: Victor Igumnov
>         Attachments: cass-scrub-threaddump.bz2
>
>
> I have noticed that when running the cleanup or scrub process on Cassandra it prevents nodetool from taking a snapshot. The snapshot process essentially blocks until the scrub or cleanup is complete. I am not exactly sure if this is intended behavior. 
> For the record, this observed behavior does not apply to repairs and major compactions.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@cassandra.apache.org
For additional commands, e-mail: commits-help@cassandra.apache.org