activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Gary Tully <gary.tu...@gmail.com>
Subject Re: KahaDB latch wait warnings
Date Wed, 08 Dec 2010 15:24:00 GMT
In the main, it is not important, that should be at debug or trace
level logging or removed altogether, it is just an indication of the
pagefile sync to disk latency and the amount of concurrent writes that
are pending at the time, an remnant of some performance tuning work
that was done for 5.4.0. the 100ms limit is arbitrary.
Do u mind tracking this with a jira issue as it will probably come up again.

On 8 December 2010 13:54, Aleksandar Ivanisevic
<aleksandar@ivanisevic.de> wrote:
>
>
> Just switched to kahadb on my amq 5.4.1 (fuse) and the log is filling
> with this:
>
>
> 2010-12-08 14:26:12,668 | WARN  | KahaDB PageFile flush: 3 queued writes, latch wait
took 119 | org.apache.kahadb.page.PageFile | ActiveMQ Journal Checkpoint Worker
> 2010-12-08 14:28:03,769 | WARN  | KahaDB PageFile flush: 7 queued writes, latch wait
took 140 | org.apache.kahadb.page.PageFile | ActiveMQ Journal Checkpoint Worker
> 2010-12-08 14:28:39,125 | WARN  | KahaDB PageFile flush: 3 queued writes, latch wait
took 112 | org.apache.kahadb.page.PageFile | ActiveMQ Journal Checkpoint Worker
> 2010-12-08 14:30:04,928 | WARN  | KahaDB PageFile flush: 8 queued writes, latch wait
took 109 | org.apache.kahadb.page.PageFile | ActiveMQ Journal Checkpoint Worker
> 2010-12-08 14:30:28,788 | WARN  | KahaDB PageFile flush: 8 queued writes, latch wait
took 18839 | org.apache.kahadb.page.PageFile | ActiveMQ Journal Checkpoint Worker
>
> quick code search turns out that this warning is fixed to 100ms
>
> http://bit.ly/gYH1Zu
>
> why 100ms and why is this important?
>
>
>



-- 
http://blog.garytully.com
http://fusesource.com

Mime
View raw message