Backup Central

Hello all. I'm 99% positive either I'm doing the configuration wrong - =
but I've read the manual and archives (maybe missing the answer) and =
have no clue as to what is going on...

I have the following setup - One machine running bacula-dir and =
bacula-sd (named backup1), another machine running bacula-sd (named =
backup2). I am using file storage on both. =20

On both are 3ware cards with disks that are allocated to the directory =
that holds the backup file storage (/bacula-nn). My plan was to use a =
pool, and have bacula send the jobs to them as determined by the =
director.

Well, everything *sort of* works. The problems I've encountered are as =
follows:

When backup1's disk space fills up, bconsole becomes practically =
unusable, as it slows down severely. During this time I see a lot of =
these messages:

19-Mar 15:54 backup1.mydomain.com: ERROR in authenticate.c:382 Unable to =
authenticate console "*UserAgent*" at client:172.16.0.134:36131.

Also, during the slowdown, mysqld is being hit extremely hard.

Now, if I switch the order of the "Storage" lines in the "Pool" section, =
and restart the director, bconsole becomes snappy again, and backups to =
the pool go to backup2. Mysqld is not taxed at all after the order is =
switched

Does any one have any idea on this? Below are the relevant =
configuration sections. I'm currently trying to restore a file from the =
'backup1' server (its disks are full), and it is being extremely =
slooooow. :)

Backup1 - /etc/bacula/bacula-dir.conf:

Storage {
Name =3D "backup1 - File"
Address =3D backup1.mydomain.com
SDPort =3D 9103
Password =3D "password"
Device =3D FileStorage
Media Type =3D File
Maximum Concurrent Jobs =3D 3
}

Storage {
Name =3D "backup2 - File"
Address =3D backup2.mydomain.com
SDPort =3D 9103
Password =3D "password"
Device =3D FileStorage2
Media Type =3D File2
Maximum Concurrent Jobs =3D 3
}

Pool {
Name =3D "File Storage"
Pool Type =3D Backup
Recycle =3D yes
AutoPrune =3D yes
Volume Retention =3D 365 days # one year
Storage =3D "backup1 - File"
Storage =3D "backup2 - File"
}

Backup1 - /etc/bacula/bacula-sd.conf:

Device {
Name =3D FileStorage
Media Type =3D File
Archive Device =3D /bacula-01
LabelMedia =3D yes; # lets Bacula label unlabeled =
media
Random Access =3D Yes;
AutomaticMount =3D yes; # when device opened, read it
RemovableMedia =3D no;
AlwaysOpen =3D no;
}

Backup2 - /etc/bacula/bacula-sd.conf:

Device {
Name =3D FileStorage2
Media Type =3D File2
Archive Device =3D /bacula-02
LabelMedia =3D yes; # lets Bacula label unlabeled =
media
Random Access =3D Yes;
AutomaticMount =3D yes; # when device opened, read it
RemovableMedia =3D no;
AlwaysOpen =3D no;
}

--=20
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.5.446 / Virus Database: 268.18.14/727 - Release Date: =
3/19/2007 11:49
=20

"James Bass" <jbass < at > hs...> wrote:

On both are 3ware cards with disks that are allocated to the directory
that holds the backup file storage (/bacula-nn).
[snip]

Also, during the slowdown, mysqld is being hit extremely hard.

Now, if I switch the order of the "Storage" lines in the "Pool" section,
and restart the director, bconsole becomes snappy again, and backups to
the pool go to backup2. Mysqld is not taxed at all after the order is
switched

Does any one have any idea on this?

I have had Bacula 1.38.x (and 1.36.x) running on a server with
a 3ware RAID controller for over a year now, and swear *at* the
3ware controller instead of swear by it like most other people.

Don't know if the thing is a pile of toss, if there's driver
problems, or if maybe the motherboard is incompatible with the
3ware card, but disk I/O actually seems to be faster through a
USB2 disk drive on the same server than it is to SATA disks on
the 3ware card! When I finally got around to hooking up a SCSI
LTO tape drive to this server last month and found the tape to be
faster as a backup media than the disks, I realized just how
bad the disk throughput really was. It seems to me like the
major problem is limited to huge sequential reads/writes like
Bacula does to monster backup files, but that's besides the point.

In any case, the description of your problem above sounds like
disk throughput problems just like I'm seeing, and if you have
30+ second delays even trying to logon to the system at the
console when things slow down then you are surely seeing the
same problem. The MySQL problem is another symptom of a disk
bottleneck, especially if that disk partition is also on the
3ware controller.

If I'm correct in my diagnosis, the explanation for your
increased performance after swapping the parameters in the
config file is because you re-started the daemons...which
stopped the file processing backlog and let the disks catch up.
Having a second storage daemon and disk controller is
probably a little bit of a plus to better disk throughput
as well.

Of course, I may be wrong, but my $0.02 is to start by
checking the disk throughput on your "backup1" box before
you dig any deeper into Bacula's configuration.....

-Arthur

-------------------------------------------------------------------------
Arthur Emerson III Email: emerson < at > ms...
Network Administrator InterNIC: AE81
Mount Saint Mary College MaBell: (845) 561-0800 Ext. 3109
330 Powell Ave. Fax: (845) 562-6762
Newburgh, NY 12550 SneakerNet: Aquinas Hall Room 6

On 3/19/07, Arthur Emerson III <emerson < at > ms...> wrote:

I have had Bacula 1.38.x (and 1.36.x) running on a server with
a 3ware RAID controller for over a year now, and swear *at* the
3ware controller instead of swear by it like most other people.

My experience is that, as long as you're not using RAID5, they're
fantastic. The onboard RAID5 implementation however performs *very*
badly. Strangely the onboard RAID50 implementation performs just as
well as all the other options (including the misnamed RAID0).

Worth doing some simple performance benchmarking (dd if=/dev/sdX
of=/dev/sdX1 count=2000000) to get a rough indicator. Also worth
watching the WIO field of top to see if you're heavily I/O bound.

--
Please keep list traffic on the list.

Rob MacGregor
Whoever fights monsters should see to it that in the process he
doesn't become a monster. Friedrich Nietzsche