cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Aleksey Yeschenko (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-11090) Hinted Handoff loop
Date Mon, 08 Feb 2016 23:46:39 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-11090?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15138023#comment-15138023
] 

Aleksey Yeschenko commented on CASSANDRA-11090:
-----------------------------------------------

I could, but it would change the semantics around closing the current writer. Right now we
only close in {{flush()}} if we are forced to switch the writer if it exceeds the max configured
hint file size. With this change, we would do it unconditionally.

> Hinted Handoff loop
> -------------------
>
>                 Key: CASSANDRA-11090
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-11090
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: Blake Eggleston
>            Assignee: Aleksey Yeschenko
>            Priority: Minor
>             Fix For: 3.0.x, 3.x
>
>
> After the hints executor finishes sending hints, another hints file is created for the
same host, which is then processed on the next go around. This continues indefinitely. The
new hint files are empty, so there shouldn't be any network traffic. However, there's still
unnecessary hint activity, and the hint crc file isn't deleted in the 3.0 branch after hints
are dispatched (but is in trunk), so any hint activity will trigger ~8640 files being created
per day until the node is restarted. Restarting the node fixes the problem, after the existing
hint files are processed. 
> This can be duplicated on cassandra-3.0 to trunk with this script: https://gist.github.com/bdeggleston/13fbb9e70c0c0bd277c7



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message