Hey Timo,
to tell you the truth, I never trusted too much vxfs. The quantity of
patches they required were enough to keep a pool of SysAdmins busy half
of the time ::).
Anyway, its true that the final performance I had lately testing vxfs
conviced me to have a go with it!
> We have splitted /var/opt/openmail/data into 10 filesystems and we do
> snapshot for all of them + /var/opt/openmail at the same time during OM
> is suspended. So the filesystems/OM database are/is consistent for the
> whole backup time, which is ( we use HP OB II) about 10 hours full and
> 30 minutes per incremental backup.
>
The information about "it is not possible to take a snapshot of several
file systems at the same time" came from HP, but apparently it's not true
anymore, and this is a good news to me, in fact mirror is a good solution
but not optimal and ... I'm moving to VSS!
> BTW, I would be interested about how long does the restoring of the OM
> database with 2M files take with ADSM. We use OB becouse I can't get it
> back fast enough with ADSM. The other reason is the number of files in
> OM server, it would just explode the ADSM database. With OB I can
> disable logging to OB database, so it will only keep track about which
> volumes were used for backup and when and how long they are protected.
>
Yep, OB-II is probably quicker when restoring, but:
1) I decided that ADSM is the only tool for back-up and I have no
intention to go back to OmniBack-II. No operator losing tapes, mixing
them up, etc.
2) ADSM Backup are quicker! I do backup all the FSs 'incremental'ly every
night, and I do circularly backup with 'select' one of these FSs
in order to reduce the fragmentation through the media of the OpenMail
pool. I did enable collocation at node level.
3) restore time ... about ((2h30m x FS) x 8FS), every FS is about 5GB.
But I'm undergoing further test in the next future ... I let you know if
you are still interested.
Ciao mauro
|