Bacula-users

Re: [Bacula-users] Bacula jobs causing VM server reboot & consequent job fail

2011-03-24 06:20:02
Subject: Re: [Bacula-users] Bacula jobs causing VM server reboot & consequent job fail
From: Marcello Romani <mromani AT ottotecnica DOT com>
To: bacula-users AT lists.sourceforge DOT net
Date: Thu, 24 Mar 2011 11:17:19 +0100
Il 24/03/2011 10.27, Sami Haahtinen ha scritto:
> On Thu, Mar 24, 2011 at 10:49, farmershort
> <bacula-forum AT backupcentral DOT com>  wrote:
>> however, I had scheduled the first big test of this new bacula system a 
>> couple of days ago - and asked it to back up about 350GB from each of 2 
>> servers - at the same time. Both of these backups should run concurrently, 
>> and back up to the same esata disk.
>>
>> I managed to get this to happen yesterday, by manually running both jobs 
>> from bconsole, but for some reason, whenever they run from schedule, then 
>> cause the VM to reboot. If you watch it happening, you see the VM go into 
>> "saving" mode, then "saved".... then some minutes later it reboots.
>
> I would start by assuming that bacula is only the trigger in the
> situation and start by running stress tests on the virtual machines.
> First for the disks and then for the network (and if you use client
> side compression, on CPU as well).
>
> It's pretty likely that the higher IO loads have caused some driver to
> fail somewhere and that is causing the situation. I'm finding it
> rather unlikely that the cause would be directly related to bacula,
> but everything is possible.
>
> Regards,

For example, VMWare 2 linux guests can have problems with virtual scsi 
disks under very high I/O loads (like those triggered by bacula 
backups), which cause the root fs to be remounted r/o, thus forcing a 
guest reset via vmware console. I've seen this happen mainly when vmware 
guest tools are installed. Suggestions on the web are to switch to a 
virutal ide controller.

This example just to say that bacula can push virtual systems to their 
limits in terms of I/O throughput, thus exposing subtle bugs that are 
seen by guest systems as hardware failures.

HTH

------------------------------------------------------------------------------
Enable your software for Intel(R) Active Management Technology to meet the
growing manageability and security demands of your customers. Businesses
are taking advantage of Intel(R) vPro (TM) technology - will your software 
be a part of the solution? Download the Intel(R) Manageability Checker 
today! http://p.sf.net/sfu/intel-dev2devmar
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users