Backup Central

Hello everybody!

I searched in the list for similar problems with backups to tape.
What I want to do is migration of jobs to tape with a Tandberg LTO-2
SCSI drive.

First of all I use bacula 2.0.2 on NetBSD 3.0.1 on a standard PC.

At the beginning, I ran the btape program and it suggested to add in my
bacula-sd.conf:

* Hardware End of Medium = No
* Fast Forward Space File = No
* BSF at EOM = yes

So it looks for the Tape Drive as below:

Device {
Name = TapeDev
Device Type = Tape
Media Type = LTO2
Archive Device = /dev/rst0
LabelMedia = yes;
Random Access = No;
# Block Positioning = no;
AutomaticMount = yes;
AlwaysOpen = yes;
RemovableMedia = yes;
Hardware End of Medium = No
Fast Forward Space File = No
BSF at EOM = yes
# Requires Mount = yes;
}

(BTW do I have to use the semicolons? They seem optional.)

Anyway, when I run btape's test again, everything seems alright, no
errors or warnings anymore.

The fill command (tested with two tapes) runs also without any problems.
It also writes about 190 GB on tape at ca 19 MB/sec.

Now, when I run a Migration Job it finds the apropriate JobIds and
starts migrating, but stops after about 17-25 GB and requires a new
Tape. Then it writes another 20 GB and again.

My Backup Jobs are stored as Volumes of 100MB to 170GB on two 500GB
harddrives first. One Volume-File per job and machine.

My Tapes are brand-new, the drive is new, too, so it should be a
hardware problem.

Here is an excerpt with the most important config-data concerning my
problem. (If I was missing something then please ask for it!) Maybe you
can help me in finding any errors?

# bacula-dir.conf:
#
# Backup Job to HD
JobDefs {
Name = "JobDef-BSD"
Type = Backup
Level = Full
FileSet = "rootdir_bsd_home"
# Schedule = "Platte1_SchedBSD"
Messages = Standard
Pool = Platte1BSD
Storage = "BackupStorageBSD"
Priority = 10
}
Job {
Name = "amelie-platte1"
Client = amelie-fd
JobDefs = "JobDef-BSD"
Schedule = "Platte1_Amelie"
Pool = Platte1-Montag
Write Bootstrap = "/backup/spool/bacula/amelie.bsr"
}
# Backup Job to Tape
JobDefs {
Name = "JobDef-BSD-Migrate"
Type = Migrate
Level = Full
Client = hugo-fd # Client with Tape-Drive
FileSet = "rootdir_bsd_home"
Selection Type = Volume
Storage = "BackupStorageBSD" # Storage with Harddisk Backup Volumes
Schedule = "Tape"
Messages = Standard
Maximum Concurrent Jobs =4
Priority = 10
}
Job {
Name = "montag-tape"
JobDefs = "JobDef-BSD-Migrate"
Pool = Platte1-Montag
Selection Type = Volume
Selection Pattern = "Platte1-Montag"
}

Pool {
Name = Platte1-Montag
Pool Type = Backup
Recycle = yes
AutoPrune = yes # Prune expired volumes
Volume Retention = 7 days
Label Format = "Platte1-Montag"
NextPool = Tape-Montag
# Storage = "BackupStorageBSD" # Now in Backup-Jobs
UseVolumeOnce = yes
}

#Migration Job on Tape
Pool {
Name = Tape-Montag
Pool Type = Backup
Storage = "BackupStorageTape"
Recycle = yes
AutoPrune = yes # Prune expired volumes
Volume Retention = 7 days
Label Format = "Montag-"
UseVolumeOnce = no
}
Storage {
Name = BackupStorageTape
Address = hugo.intra
SDPort = 9103
Password = "StPass"
Device = TapeDev
Media Type = LTO2
}

Another strange phenomena I would like to ask you is what happens with a
job that has been only migrated partially (i.e. stopped with an
error/was aborted by me)? As I noticed, bacula won't try to migrate it
again?! Is the job lost then?

Thank you very much for any hints solving my problem!

Have a nice day,
Damian

Damian Lubosch wrote:
Hello everybody!

I searched in the list for similar problems with backups to tape.
What I want to do is migration of jobs to tape with a Tandberg LTO-2
SCSI drive.

Ok, I think that I had thermal problems. Keep your Tape Drives cool ;-)

But now I noticed another problem: during a migration job my drive stops
quite often. The migration job finished with that message:

14-Mar 13:57 hugo-dir: The following 1 JobIds will be migrated: 3
14-Mar 13:57 hugo-dir: Migration using JobId=3 Job=zoii.2007-03-13_21.21.09
14-Mar 13:57 hugo-dir: Bootstrap records written to /var/spool/bacula/hugo-dir.restore.1.bsr
14-Mar 13:57 hugo-dir:
14-Mar 13:57 hugo-dir: The job will require the following
Volume(s) Storage(s) SD Device(s)
===========================================================================
14-Mar 13:57 hugo-dir:
14-Mar 13:57 hugo-dir: Platte1-Freitag0001 BackupStorageWin BackupDev1
14-Mar 13:57 hugo-dir:
14-Mar 13:57 hugo-dir: Start Migration JobId 10, Job=freitag-tape.2007-03-14_13.57.54
14-Mar 13:58 hugo-sd: Ready to read from volume "Platte1-Freitag0001" on device "BackupDev1" (/backup/bacula-backups).
14-Mar 13:58 hugo-sd: Wrote label to prelabeled Volume "Freitag" on device "TapeDev" (/dev/rst0)
14-Mar 13:58 hugo-sd: Forward spacing Volume "Platte1-Freitag0001" to file:block 0:209.
14-Mar 18:21 hugo-sd: End of Volume at file 39 on device "BackupDev1" (/backup/bacula-backups), Volume "Platte1-Freitag0001"
14-Mar 18:21 hugo-sd: End of all volumes.
14-Mar 18:21 hugo-dir: Bacula 2.0.2 (28Jan07): 14-Mar-2007 18:21:14
Prev Backup JobId: 3
New Backup JobId: 11
Migration JobId: 10
Migration Job: freitag-tape.2007-03-14_13.57.54
Backup Level: Full
Client: hugo-fd
FileSet: "rootdir_bsd_home" 2007-03-13 21:20:21
Read Pool: "Platte1-Freitag" (From Job resource)
Read Storage: "BackupStorageWin" (From user selection)
Write Pool: "Tape-Freitag" (From Job Pool's NextPool resource)
Write Storage: "BackupStorageTape" (From Storage from Pool's NextPool resource)
Start time: 14-Mar-2007 13:58:00
End time: 14-Mar-2007 18:21:14
Elapsed time: 4 hours 23 mins 14 secs
Priority: 10
SD Files Written: 2,159,801
SD Bytes Written: 168,088,680,253 (168.0 GB)
Rate: 10642.6 KB/s
Volume name(s): Freitag
Volume Session Id: 1
Volume Session Time: 1173877051
Last Volume Bytes: 168,285,164,544 (168.2 GB)
SD Errors: 0
SD termination status: OK
Termination: Migration OK

Sometimes the drive ran like for 10 minutes or so without stoping, sometimes it stopped and rewound every ten seconds.

The transfer rate should also be 20MB/s or more, since compression is enabled but because of the stopping it dropped to 10MB/s.

I am using a standard PC with 2 recent WesternDigital IDE-ATA harddrives, my Tapedrive (Tandberg LTO2) is connected to a SCSI Controller, the computer has 512 MB RAM, and a 2.8GHz Celeron CPU. During the migration no other jobs are running on the machine.

Sometimes I noticed high database (PostgreSQL) activity. Could it have to do with the 2.1 million files for backup?

Do you experience similar behavior? I am afraid that this could shorten my drives lifespan
dramastically.

Are there any setting to speed up the migration?

Thank you very much in advance,
Damian Lubosch