Re: [Bacula-users] parallelizing ClientRunBeforeJobs

[Dan L. <da...@la...>]

On 2013-01-17 05:34, lst...@kw... wrote:
> Zitat von Tilman Schmidt <t.s...@ph...>:
>
>> One of my Bacula servers backs up 18 clients. Many of these
>> have a lengthy ClientRunBeforeJob. (SVN hot backups, database
>> dumps, Windows system state backups - you get the picture.)
>> As a consequence, more than half of the elapsed time of my
>> nightly backup runs is actually spent waiting for one of the
>> clients to complete its Before job while everything else sits
>> idle.
>>
>> How are others handling this situation? Any smart ideas how
>> to get the before jobs to run in parallel, preferably keeping
>> the actual writing to tape sequential?
>
> Hello
>
> we use the spooling feature and run all *jobs* concurrently. With 
> this
> you get adjustable sized "blocks" written to tape sequential to
> maximize throughput and concurrent spooling for all other jobs. Works
> trouble free but you need a really fast spool device for anything
> faster than LTO-3.

I'm not sure how useful this is.  But perhaps admin jobs will be hepful
for this.  They aren't run before, but they don't backup, they just 
run.

-- 
Dan Langille - http://langille.org/