Re: [Bacula-users] multiple concurrent jobs issue

[Simon E. <si...@si...>]

Hi,

[resend to the list due to wrong sender address, sorry if this arrives 
twice]

Kern Sibbald skrev:
> Hello,
> 
>>From what you have described here, it appears that Bacula is recycling a 
> volume while it is being used or has been reserved for use.  I am not sure 
> what is going on for the following reasons:
 >
> 1. You speak of rotation of volumes, but there is no such concept in Bacula.  
> I am assuming you mean recycling of volumes.  Please try to use the right 
> terminology as it helps ensure proper understanding.

Sorry about the lack of proper terminology. No, I don't mean recycling, 
I mean when bacula switches from one Volume in a pool to the next, ie. 
what happens when 'Volume Use Duration' is reached for a volume. I'm not 
sure if that particular concept has a name.

> 2. It appears that you are using Volume retention periods to attempt to force 
> Bacula to use a different volume.  This absolutely will not work.  It is 
> probably the main reason you are having problems.

Sorry for the misunderstanding, no that's not what I'm trying to do.
A typical sample pool definition we're using:

Pool {
   Name = Default-Windows-94
   Pool Type = Backup
   Recycle = yes
   AutoPrune = yes
   Volume Retention = 7 days
   Volume Use Duration = 23h
   LabelFormat = "Vol-windows-94"
}

The problem we're encountering is when:
* Job A starts running.
* 'Volume Use Duration' is reached.
* Job B starts running but fails due to A still running and writing to 
the old volume, preventing the new volume from being used.
* Job C, D, E have the same problem as B.
* Job A finishes.
* Job F starts running, the new volume is used since the old volume is 
no longer being written to, so job G succeeds.
* Job G, H, I succeed.

> 3. Normally if Bacula recycles a volume while jobs are using it, you have an 
> important configuration inconsistency (retention periods, not setting 
> reasonable volume sized, ...)

Yepp, I understand that.

> 4. A proper solution to your problem is to set explicit maximum volume sizes 
> and then ensure that your volume retention period will not occur while jobs 
> are writing to the volume.

As far as i can tell we've got that much covered with the 'volume use 
duration' and 'volume retention' settings in the pool above. If full 
configuration files would help I'll be happy to anonymize them where 
necessary and post them tomorrow.

Regards,
-- 
Simon Ekstrand