Re: how to recover the whole month backups
2007-12-07 22:01:09
Ahh, so you have a 1TB RAID on which you have set up 30 virtual tapes,
and those 30 tapes occupy 300GB after running for 30 days. You could
add more virtual tapes as suggested by Jean-Louis, however, at
300GB/month you are going to run out of space sometime after 3 months.
You're also putting all your eggs in one RAID, to twist a phrase. If
that were a server class RAID built with server class SCSI drives, I
might trust it; but, if it is cheap hardware SATA type stuff, I've
heard more than a few horror stories of multiple drive failures
leading to data loss.
I know ur concern. I am using SATA2 HDD 7200rpm RAID 5 (h/w raid). If
HDD failure I can still rebuild it.
So you have a couple of issues. You should think about redundancy and
how well you are covering yourself. This is strongly influenced by
your budget, but at least consider having a hot spare configured in
your RAID. And, be sure to keep an eye on things. Watch log files.
Perhaps implement some automated checking and notification of hardware
errors. That's one reason I like real tapes -- the possibilities for
catastrophic data loss are almost nil.
Real tapes is better but budget is limited to me to have it. It is not
cheap. That's why I am using tapeless. Furthermore, who are going to
load tapes if it is full let say in midnight/early morning.
Your other issue is bringing your storage capacity and your backup
policies into line with one another. You indicate you want coverage
for a year. Using your current configuration and expanding the number
of virtual tapes will have you running out of storage capacity in just
over 3 months. Your daily needs aren't that big. I don't know what
your configuration is (dump cycle, runs per cycle, etc.), but
adjusting those probably wouldn't help enough, since you need 3 or 4
times as much storage capacity. However, since your daily backups are
not too big, you could perhaps once a month force all your DLEs to
full and then mark that tape as no-reuse. At the end of some period,
you'll have a bunch of full backups marked no-reuse and a full cycle
of vtapes that just keep getting turned over (maybe a 6 week cycle?).
You will have to determine what fits into your storage capacity. Then
you could just remove the no-reuse tag for the oldest and let it go
back into the cycle. At that point you are in a pattern of each month
or so forcing a full, marking it no-reuse, and releasing the oldest
no-reuse to go back into the cycle. Your cycle would be totally
automated by Amanda, but your forced fulls would be manual (though you
could script it). That's just one possibility.
Yeah, that is better solution. I have never thought about it. I will try
to apply it.
FYI, my dumpcycle is 30 days, tapecycle 31 tapes and no runspercycle
settings.
I wouldn't get into multiple configurations until you are really
comfortable with it. Otherwise, it could make things too complicated
as well as less efficient for your storage capacity. Better to have
all your backups and indexes under one configuration. Amanda can
continue to plan incrementals based on the no-reuse fulls since they
are part of the same configuration. But that's just my opinion for
your situation at this point in time.
Much thanks for ur meaningful explanations. It will save disk space if I
mark my tapes to no-reuse and I wouldn't waste my times to restore.
zuki.vcf
Description: Vcard
|
|
|