Last week I upgraded my Bacula install to 2.0.2 without any issues during the
upgrade process.
All seemed to be working well until Friday, but that was only because I hadn't
discovered this issue yet.
I run several jobs each night, with Full backups on Friday night and
incrementals every other night. Once the backups are completed I have bacula
setup to verify a few of these tape jobs against the catalog with Volume to
Catalog jobs.
These verify jobs have a higher priority, so they shouldn't start until after
the backup jobs are complete. I also schedule them to start 10 minutes after
the backups at 23:15. (All backups are scheduled to start at 23:05)
This behaviour seems to have changed in Bacula 2.0.2 however. They seem to
launch immediately and select which tape they'll use for the verify.
Here's a log excerpt from a job running under 1.38.5 (My old version)
02-Mar 01:09 backup1-dir: Verifying against JobId=6017 Job=fs1.2007-03-01_23.00.00
02-Mar 01:09 backup1-dir: Bootstrap records written to /export/bacula/var/backup1-dir.restore.Verify-fs1.2007-03-01_23.15.00.bsr
02-Mar 01:09 backup1-dir:
02-Mar 01:09 backup1-dir: The job will require the following Volumes:
02-Mar 01:09 backup1-dir:
02-Mar 01:09 backup1-dir: 000009
02-Mar 01:09 backup1-dir:
Even tho it was scheduled to run at 23:15, it didn't start until 1:09 the next
morning. This is the expected behaviour.
Here's a log excerpt from Friday night under 2.0.2:
09-Mar 23:15 backup1-dir: Verifying against JobId=6101 Job=fs1.2007-03-08_23.00.00
09-Mar 23:15 backup1-dir: Bootstrap records written to /export/bacula/var/backup1-dir.restore.11.bsr
09-Mar 23:15 backup1-dir:
09-Mar 23:15 backup1-dir: The job will require the following
Volume(s) Storage(s) SD Device(s)
===========================================================================
09-Mar 23:15 backup1-dir:
09-Mar 23:15 backup1-dir: 000010 Tape OfficeAutochanger
09-Mar 23:15 backup1-dir:
10-Mar 20:57 backup1-dir: Start Verify JobId=6125 Level=VolumeToCatalog Job=Verify-fs1.2007-03-09_23.15.00
As can be seen, the job started at 23:15, selected tape '000010' and then
waited until the next night to run (Full backups take most of a day
)
Meanwhile, the full backup ran, writing to 2 other tapes.
The problem encountered by bacula is that tape 000010 had been removed from the
Autochanger on Friday morning and replaced with this weeks incremental tape. So
Bacula blocked, waiting for a tape, until I discovered and cancelled this job
this morning. Not ideal, but it did bring this issue to light.
What I'd like to know is if this was a deliberately designed feature, or if it
classifies as a bug? If it's deliberate, is there anyway to control when it
performs it's tape selection, or even which job it verify? I've scanned the
Jobs section of the latest manual but didn't see anything obvious.
For now I can either work around it(With a Run script running the verify job),
or even work with it (I've wanted to run the Full Backup job verifies on Sunday
for a while anyways.)
Cheers,
Troy Daniels
Systems Administrator
iTouch Australia (pty) ltd.
