[Linux-cluster] fencing issue in 2 nodes cluster

Digimer lists at alteeve.ca
Fri Aug 3 14:49:24 UTC 2012


That should like it's fine then.

As a general rule though, I recommend roughly a 2:1 test:implement 
ratio. So if it takes you a week to setup your cluster, take two weeks 
to think of and test every failure you can think of. It's amazing how 
many corner cases you find this way (and much of the two weeks will 
actually be improving your design).

digimer

On 08/03/2012 04:13 AM, AKIN ÿffffffffffd6ZTOPUZ wrote:
> dont take any error
> and yes removed the method from command  (default is onoff)
> exact  line is at below:
> <fencedevice agent="fence_ipmilan" ipaddr="192.168.11.68" lanplus="1"
> login="clsfenceadmin"  name="fence_node1" passwd="clsfenceadmin"
> power_wait="4" verbose="true"/>
> and I tested fencing using manually (making a kernel crash:Digimer
> note)   and   command with fence_node node1 .
> İt worked and services is relocated properly
> thanks
>
> *From:* Heiko Nardmann <heiko.nardmann at itechnical.de>
> *To:* linux-cluster at redhat.com
> *Sent:* Friday, August 3, 2012 10:40 AM
> *Subject:* Re: [Linux-cluster] fencing issue in 2 nodes cluster
>
> Am 03.08.2012 09:12, schrieb AKIN ÿffffffffffd6ZTOPUZ:
>> Hi
>> thanks all repilies
>>  I added verbose parameter into fence agent line in cluster.conf
>> according to  Heiko (verbose="true",method="onoff")
>> and I read the web link sent by Digimer .I tested fencing using
>>   "echo c > /proc/sysrq-trigger" on  node1.
>> after that I saw fencing occured and related services    moved to
>> other node properly.
>> in this case    can we say , fencing is ok ?    or also should I  use
>> fence_node command   on node   for being sure?
>>
>
> No errors anymore? Did you reduce the number of parameters as Digimer
> suggested? Then it would be maybe interesting for others to see your
> working fence configuration as configured inside cluster.conf.
>
> I would suggest to test a little bit too much than doing not enough
> testing ... ;-) ... debugging afterwards is more costly than doing
> testing before ...
>
> So my recommendation would be to test that manually started fencing,
> too, to see whether any messages/errors appear.
>
> After that you are probably fine ...
>
>
> Kind regards,
>
>      Heiko
>
>
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com <mailto:Linux-cluster at redhat.com>
> https://www.redhat.com/mailman/listinfo/linux-cluster
>
>
>
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster
>


-- 
Digimer
Papers and Projects: https://alteeve.com




More information about the Linux-cluster mailing list