>>>>> On Fri, 07 Dec 2007 10:02:58 +0000, Damian Brasher said:
> 
> Martin Simmons wrote:
> 
> > Also, you could attach gdb to each daemon and run the gdb command
> >
> > thread apply all bt
> >
> 
> 11th Nov 07-----------------------------------------------------------
> Have attached /sbin/bacual-dir /sbin/bacula-fd and /sbin/bacula-sd to
> gdb, run the commands
> and will now wait until the error condition repeats.
> 
> Will post the output to [(gdb)info file] and [(gdb)thread apply all bt]
> as soon as I have the error condition
> as well as the dir/fd and sd status.
> ----------------------------------------------------------------------
> 
> The error has occurred again. I decided to start bacula with the init 
> scripts and attached gdb to the running process.
> 
> All the details descibing this problem are at the beginning of the thread.
> 
> As the job halted this was the output from the bconsole, about 200 lines 
> of roughly the same as below:-
> 
> ...Orphaned buffer:  backup-dir      8 bytes buf=9e1f010 allocated at 
> workq.c:167
> Orphaned buffer:  backup-dir     16 bytes buf=9e1eee0 allocated at jcr.c:247
> Orphaned buffer:  backup-dir    528 bytes buf=9e1f038 allocated at jcr.c:255
> Orphaned buffer:  backup-dir    528 bytes buf=9e23ab8 allocated at job.c:953
> Orphaned buffer:  backup-dir    528 bytes buf=9e23ce8 allocated at 
> job.c:1130
> Orphaned buffer:  backup-dir      6 bytes buf=9e23f50 allocated at 
> ua_server.c:105
> Orphaned buffer:  backup-dir    316 bytes buf=9e23f78 allocated at 
> ua_server.c:192
> Orphaned buffer:  backup-dir    804 bytes buf=9e24338 allocated at 
> bsock.c:429
> Orphaned buffer:  backup-dir    707 bytes buf=9e24c40 allocated at 
> mem_pool.c:198
> Orphaned buffer:  backup-dir    707 bytes buf=9e24680 allocated at 
> mem_pool.c:198
> Orphaned buffer:  backup-dir     24 bytes buf=9e1f268 allocated at 
> job.c:1153
> Orphaned buffer:  backup-dir     40 bytes buf=9e1f2a0 allocated at 
> alist.c:53...

That is very unexpected.  The only time I've seen 'Orphaned buffer' messages
is after killing the Director.  Could that have happened?  Were there any
other messages in the log when the job halted?


> 
>  From command: status all the only unusual output is:
> 
> Running Jobs:
> JobId 166 Job holly.2007-12-06_23.25.09 is running.
>      Backup Job started: 07-Dec-07 01:41
>      Files=50,030 Bytes=12,066,088,479 Bytes/sec=407,211
>      Files Examined=66,825
>      Processing file: /etc/httpd/conf/httpd.conf
>      SDReadSeqNo=6 fd=7
> Director connected at: 07-Dec-07 09:55
> 
> The sd status is:
> 
> backup-sd Version: 2.2.5 (09 October 2007) i686-pc-linux-gnu redhat 
> Enterprise release
> Daemon started 06-Dec-07 11:42, 4 Jobs run since started.
>   Heap: heap=217,088 smbytes=160,745 max_bytes=161,943 bufs=124 max_bufs=133
> Sizes: boffset_t=8 size_t=4 int32_t=4 int64_t=8
> 
> Running Jobs:
> Writing: Full Backup job holly JobId=166 Volume="Thursday1"
>      pool="Thursday" device="LTO-2" (/dev/nst0)
>      spooling=0 despooling=0 despool_wait=0
>      Files=50,030 Bytes=12,073,480,104 Bytes/sec=404,607
>      FDReadSeqNo=646,525 in_msg=552497 out_msg=6 fd=8

What did the client status show?

>From the FD backtraces, it looks like either the job has finished cleanly or
the thread running the job has evaporated without trace.

>From the SD backtraces, it looks like the SD is waiting for the FD to confirm
that the job has finished.

Was the gdb attached to the FD while the job was running or did you attach it
after it started hanging?  If the latter, are you 100% sure that the bacula-fd
process was not restarted somehow?

Do netstat or lsof show any socket connections between the SD and the FD when
the job has reached this hanging point?

I think you might have to run the SD (and possibly the FD) at debug level 200
to collect info about what happens at the end of the job.

__Martin

-------------------------------------------------------------------------
SF.Net email is sponsored by:
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users

Reply via email to