[prev in list] [next in list] [prev in thread] [next in thread] 

List:       redhat-linux-cluster
Subject:    Re: [Linux-cluster] Using ping as a heuristic not a great idea?
From:       Don Hoover <dxh () yahoo ! com>
Date:       2010-05-21 15:08:27
Message-ID: 838401.93092.qm () web65508 ! mail ! ac4 ! yahoo ! com
[Download RAW message or body]

Well, we seem to have found something that works even when running the heartbeats \
across the production network.

Using this we were able to maintain heartbeats as well as have the quorum ping \
heuristic work even when we flooded the interface on the node.


Instead of the often cited "ping -c1 t1 <router>", we used "ping -c3 -t3 -W1", as \
well as upping the heuristic interval to 10, and keeping the quorum interval at 5 it \
worked during our testing of flooding the interface to get it to fail quorum.   With \
the standard c1 t1 it was easy to get quorum to fail just by doing a couple of big \
ftp jobs on the node.  We had the interface pegged at pretty much the theoretical max \
throughput and were still able to keep getting the ping heuristic to work with these \
settings.


Here is an example showing what we used and finally settled on:

<quorumd interval="5" label="rhcsqdisk" min_score="1" tko="10" votes="1">
     <heuristic interval="10" program="/bin/ping -c3 -t3 -W1 192.168.0.1" score="1"/>
</quorumd>


Note, since we bumped up our quorumd interval from 2 to 5 (seconds between writes to \
qdisk), we had to up the totem token to "100000" (100 seconds = 2 x qdisk interval x \
qdisk tko).

--
Linux-cluster mailing list
Linux-cluster@redhat.com
https://www.redhat.com/mailman/listinfo/linux-cluster


[prev in list] [next in list] [prev in thread] [next in thread]