Hi,
just a hint about "sub-optimal" behavior of bacula's storage daemon.
There's an autochanger with two LTO-3 tape drives. Sometimes, starting
some (mostly identical) jobs in parallel leads to an obscure situation:
Although all started jobs go into the same pool and on the same tape,
the storage daemon attaches the jobs to different tape drives. Of
course, the tape can be loaded into one drive only at the same time.
Consequently, the second drive is blocked, waiting for the tape that's
already in use in the first drive, and some jobs are waiting there
without any need.
Device status:
Autochanger "Tape1" with devices:
"Tape1Left" (/dev/rmt/0cbn)
"Tape1Right" (/dev/rmt/1cbn)
Device "Disk1" (/var/services/bacula/store1) is not open or does not exist.
Device "Tape1Left" (/dev/rmt/0cbn) is mounted with Volume="000000L1" Pool="Test"
Slot 1 is loaded in drive 0.
Total Bytes=31,195,990,677 Blocks=483,609 Bytes/block=64,506
Positioned at File=58 Block=876
Device "Tape1Right" (/dev/rmt/1cbn) is not open or does not exist.
Device is BLOCKED waiting for media.
Drive 1 is not loaded.
====
In Use Volume status:
000000L1 on device "Tape1Left" (/dev/rmt/0cbn)
====
...
Running Jobs:
Writing: Incremental Backup job isg-201 JobId=1335 Volume="000000L1"
pool="Test" device=""Tape1Left" (/dev/rmt/0cbn)" <<<<<<<<<<<<<<<<<< !!!
Files=0 Bytes=0 Bytes/sec=0
FDReadSeqNo=6 in_msg=6 out_msg=4 fd=9
Writing: Incremental Backup job isg-209 JobId=1337 Volume="000000L1"
pool="Test" device=""Tape1Left" (/dev/rmt/0cbn)"
Files=0 Bytes=0 Bytes/sec=0
FDReadSeqNo=6 in_msg=6 out_msg=4 fd=11
Writing: Incremental Backup job isg-206 JobId=1336 Volume="000000L1"
pool="Test" device=""Tape1Left" (/dev/rmt/0cbn)"
Files=0 Bytes=0 Bytes/sec=0
FDReadSeqNo=6 in_msg=6 out_msg=4 fd=13
Writing: Incremental Backup job lns-104 JobId=1338 Volume="000000L1"
pool="Test" device=""Tape1Right" (/dev/rmt/1cbn)" <<<<<<<<<<<<<<<<< !!!
Files=0 Bytes=0 Bytes/sec=0
FDReadSeqNo=6 in_msg=6 out_msg=4 fd=12
====
As you can see, the tape is loaded in the left drive (0), and the jobs
which are attached to this drive, can start their backup sessions. The
other job has to wait for the same tape to be loaded into drive 1,
and generate messages like
Subject: Bacula: Intervention needed for lns-104.2006-05-09_09.45.03
...
09-May 09:49 bup-serv-sd: Please mount Volume "000000L1" on Storage Device
"Tape1Right" (/dev/rmt/1cbn) for Job lns-104.2006-05-09_09.45.03
This is no hard problem, because, after terminating all jobs using drive 0,
the tape is unmounted, unloaded from drive 0 to magazine, loaded again
into drive 1, and the lost job runs successfully.
But, this behaviour might produce problems, f.i. if there is enough
system's capacity to run all jobs in parallel (as is it in my case),
and some jobs, waiting needlessly for the other drive, would run out
of max waiting time.
And, as told above, this behaviour blocks the second drive as long
as the jobs on the first drive run, needlessly.
Might be reasons to improve this in a future release of bacula.
Best regards,
Robert
--------------------------------------------------------
++ German Research Center for Artificial Intelligence ++
--------------------------------------------------------
Robert Wirth, Stuhlsatzenhausweg 3, D-66123 Saarbruecken
@office: +49-681-302-5078/5572 ++ @fax: +49-681-302-5341
mailto:[EMAIL PROTECTED] ++ http://www.dfki.de/~wirth
--------------------------------------------------------
-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users
