On Tue, Jan 26, 2010 at 01:08:16PM +0100, Kern Sibbald wrote: > On Tuesday 26 January 2010 12:55:25 Graham Keeling wrote: > > On Tue, Jan 26, 2010 at 11:25:44AM +0000, Graham Keeling wrote: > > > So, I have a patch that seemed to make it do the intended thing > > > (attached), although there seem to more problems. Another report to > > > come... > > > > Here is the problem I mentioned (or possibly two problems). > > > > I'm using disk-based backups, one job per volume. > > Full backups have their own pool. > > Incrementals have their own pool. > > RecyclePool and ScratchPool on both point to the same Scratch pool, called > > "Disk 1". > > > > > > a) I run a Full backup (creating backup-0001). > > b) I run an Incremental backup (creating backup-0002). > > > > tserv volumes # ls -l > > total 2920 > > -rw-r----- 1 admin-bacula-sd admin-bacula-sd 2416814 Jan 26 11:37 > > backup-0001 -rw-r----- 1 admin-bacula-sd admin-bacula-sd 557692 Jan 26 > > 11:37 backup-0002 > > > > c) I purge backup-0002. > > > > tserv volumes # ls -l > > total 2372 > > -rw-r----- 1 admin-bacula-sd admin-bacula-sd 2416814 Jan 26 11:37 > > backup-0001 -rw-r----- 1 admin-bacula-sd admin-bacula-sd 93 Jan 26 > > 11:38 backup-0002 > > > > Good, this looks like it worked. > > > > > > d) I purge backup-0001. > > > > Enter *MediaId or Volume name: *1 > > 1 File on Volume "backup-0001" purged from catalog. > > There are no more Jobs associated with Volume "backup-0001". Marking it > > purged. Connecting to Storage daemon Disk 1 at 192.168.100.121:9103 ... > > Recycled volume "backup-0001" on device "Disk 1.0" (/write/mnt/Disk > > 1/var/lib/bacula/volumes), all previous data lost. 3917 Volume recycled > > New Pool is: Disk 1 > > *q > > tserv volumes # ls -l > > total 2372 > > -rw-r----- 1 admin-bacula-sd admin-bacula-sd 2416814 Jan 26 11:37 > > backup-0001 -rw-r----- 1 admin-bacula-sd admin-bacula-sd 93 Jan 26 > > 11:38 backup-0002 > > > > PROBLEM 1: I've now purged backup-0001, but it has not been truncated. > > (maybe it truncated backup-0002 instead of backup-0001?) > > > > > > e) I do another backup. It seems to complete OK. > > > > Looking back in my log, it appears to have chosen backup-0001. > > > > Jan 26 11:42:33 bacula-dir: tserv-dir JobId 3: No prior Full backup Job > > record found. Jan 26 11:42:33 bacula-dir: tserv-dir JobId 3: No prior or > > suitable Full backup found in catalog. Doing FULL backup. Jan 26 11:42:35 > > bacula-dir: tserv-dir JobId 3: Start Backup JobId 3, > > Job=NetPilots:BackupPilot.2010-01-26_11.42.33_09 Jan 26 11:42:35 > > bacula-dir: tserv-dir JobId 3: Recycled volume "backup-0001" Jan 26 > > 11:42:35 bacula-dir: tserv-dir JobId 3: Using Volume "backup-0001" from > > 'Scratch' pool. ... > > > > But, on the disk, bytes have been written to backup-0002! > > tserv volumes # ls -l > > total 4712 > > -rw-r----- 1 admin-bacula-sd admin-bacula-sd 2416814 Jan 26 11:37 > > backup-0001 -rw-r----- 1 admin-bacula-sd admin-bacula-sd 2395418 Jan 26 > > 11:42 backup-0002 > > > > I can use vi to look at backup-0002, and the contents say that it is > > actually backup-0001! > > > > PROBLEM 2: Data written to disk volumes ends up in the wrong file. > > > > > > I now have two volumes that think they are backup-0001. > > > > >From this point onwards, a kind of chaos happens. > > As you said: "Oh dear!" we have a problem.
Also, I just did the same test without the 'Action On Purge = Truncate' option, and it didn't write to the wrong place. ------------------------------------------------------------------------------ The Planet: dedicated and managed hosting, cloud storage, colocation Stay online with enterprise data centers and the best network in the business Choose flexible plans and management services without long-term contracts Personal 24x7 support from experience hosting pros just a phone call away. http://p.sf.net/sfu/theplanet-com _______________________________________________ Bacula-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/bacula-devel
