Bacula-users

Re: [Bacula-users] SD crashes

2012-02-16 15:35:06
Subject: Re: [Bacula-users] SD crashes
From: Martin Simmons <martin AT lispworks DOT com>
To: bacula-users AT lists.sourceforge DOT net
Date: Thu, 16 Feb 2012 17:25:54 GMT
>>>>> On Wed, 15 Feb 2012 17:46:33 +0000, Joe Nyland said:
> 
> On 15 Feb 2012, at 16:52, Martin Simmons wrote:
> 
> >>>>>> On Wed, 15 Feb 2012 13:31:10 +0000 (GMT), Joe Nyland said:
> >> 
> >> On 14 Feb, 2012,at 03:47 PM, Martin Simmons <martin AT lispworks DOT com> 
> >> wrote:
> >> 
> >>>>>>>> On Tue, 14 Feb 2012 10:34:31 -0500, John Drescher said:
> >>>> 
>>>>> I've been running the SD using the following command (I know the 
>>>>> combination
>>>>> of options I have used may be excessive, but I wanted as much chance of
>>>>> catching the error as I could!) since yesterday afternoon:
>>>>>   sudo bacula-sd -c /etc/bacula/bacula-sd.conf -d 100 -dt -f -u bacula -g
>>>>> tape -m -v | tee -a /mnt/array/bacula-sd.screen.log
>>>>> 
>>>>> However, (as luck would have it) I've not seen the behaviour I originally
>>>>> reported whilst running with debug options.
>>>>> 
>>>>> Is there any way in which running the SD with the combination of options I
>>>>> have used above, could cause any different behaviour of the SD? Or 
>>>>> interfere
>>>>> in any way with it? I'm asking, becuase I have re-enabled all of the 
>>>>> backups
>>>>> jobs I have on the server, and I have still not seen it crash again.
>>>>> 
> >>>> 
> >>>> Could be a timing issue that the delay in writing the log causes the
> >>>> bad behavior to not happen.. Those types of problems are hard to
> >>>> debug.
> >>> 
> >>> Running it under gdb without the debug options is better approach in that
> >>> case.
> >>> 
> >>> http://www.bacula.org/5.2.x-manuals/en/problems/problems/What_Do_When_Bacula.html#SECTION00640000000000000000
> >>> 
> >>> When it 'crashes' (though it sounds more like 'hangs' is a better word),
> >>> interrupt gdb with Ctrl-c to get back to the gdb shell window (as in step 
> >>> 8).
> >>> 
> >>> __Martin
> >>> 
> >>> ------------------------------------------------------------------------------
> >>> Keep Your Developer Skills Current with LearnDevNow!
> >>> The most comprehensive online learning library for Microsoft developers
> >>> is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
> >>> Metro Style Apps, more. Free future releases when you subscribe now!
> >>> http://p.sf.net/sfu/learndevnow-d2d
> >>> _______________________________________________
> >>> Bacula-users mailing list
> >>> Bacula-users AT lists.sourceforge DOT net
> >>> https://lists.sourceforge.net/lists/listinfo/bacula-users
> >>> 
> >> 
> >> 
> >> Martin and John, thank you for your replies.
> >> 
> >> I since yesterday afternoon, bacula-sd has been running under gdb using 
> >> the instructions in the manual for my Bacula version, however I've still 
> >> not seen the issue originally reported.
> >> 
> >> I agree with you both that by running it under a debugging process, it 
> >> seems a delay is introduced which is suppressing the error in some way. Is 
> >> my best bet just to leave bacula-sd running under gdb and hope that my 
> >> full backups over the weekend may highlight the issue? Or is there another 
> >> way I could debug this?
> > 
> > Assuming the error causes the SD to hang (rather than exit), then you could
> > run it without gdb and then attach gdb to it when it hangs (use gdb -p 
> > $pid).
> > 
> > __Martin
> > 
> > ------------------------------------------------------------------------------
> > Virtualization & Cloud Management Using Capacity Planning
> > Cloud computing makes use of virtualization - but cloud computing 
> > also focuses on allowing computing to be delivered as a service.
> > http://www.accelacomm.com/jaw/sfnl/114/51521223/
> > _______________________________________________
> > Bacula-users mailing list
> > Bacula-users AT lists.sourceforge DOT net
> > https://lists.sourceforge.net/lists/listinfo/bacula-users
> 
> Ok, that sounds reasonable.
> 
> One question, does bacula-sd need to be running with "-s              no 
> signals (for debugging)" or will gdb be able to provide enough info without 
> this option? (My default options on Ubuntu 10.04 are: -c <config file> -u 
> <user> -g <group>)
> 
> Would using the -s option introduce the sort of delay we mentioned earlier, 
> and thus limit my changes of reproducing the issue?

I don't think -s will have any effect on delays, but you shouldn't need it if
you attach gdb to the hanging process.

__Martin

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing 
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users

<Prev in Thread] Current Thread [Next in Thread>