[Linux-cluster] scsi_watchdog

[Ryan O'Hara]

isplist at logicore.net wrote:
> The scsi_watchdog.conf was added to all of my nodes when I did a yum update on 
> all nodes connected to this cluster. I'm not even sure why it's in the 
> /etc/cluster directory and I don't use standard SCSI since I'm using fibre 
> channel. 
> 
> I'm now suspecting that what ever this is, it is the cause of my web server 
> slow down's. I can't seem to find much about this so thought I should ask.

scsi_watchdog.conf is only being used if 1) you use scsi reservations as 
a fence method and 2) you explicitly state in a sysconfig file that you 
want scsi_watchdog to run. I'm guessing that neither of these apply to 
use, and therefore the scsi_watchdog is having no effect. It is disabled 
by default.

> On Wed, 21 Nov 2007 10:31:32 -0600, Ryan O'Hara wrote:
>> isplist at logicore.net wrote:
>>
>>> I didn't notice that something had changed until now after I upgraded all
>>> of
>>> the nodes. I see in /etc/cluster the following;
>>>
>>> -rw-r-----   1 root root  1249 Nov 10 18:40 .cluster.conf
>>> -rw-r-----   1 root root  1249 Nov 13 09:50 cluster.conf
>>> -rw-r--r--   1 root root    34 Aug 30 17:55 scsi_watchdog.conf
>>>
>>> Since that update, the web nodes have been VERY slow, sluggish, barely
>>> responding at times to web requests.
>>>
>>> Could that be related to this? I am not finding a whole lot about this  
>>> on the
>>> net?
>>>
>> What changed?
>>
>> scsi_watchdog.conf should have no effect. It isn't even uses unless you
>> configure watchdog (softdog) support for scsi reservations.
>>
>> Ryan
> 
> 
> 
> 
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster