|
Re: how to recover the whole month backups
2007-12-05 10:21:56
fedora wrote:
> (hoogendyk wrote:)
yup. I would say you are a bit confused.
The 26th is the last full backup (the 0 in the column after the date),
and the 29th is the most recent incremental based on that full (the 1 in
the column after the date). Anything else is older, out of date,
redundant . . . whatever you want to call it. If you had messed up a
file and then it got backed up, you might want to go to an older one.
What, exactly, is it that you are trying to do? And Why?
Actually I backed up one month (30 tapes). At the end of month tapes will be
overwritten. So, I want to restore for the whole month before tapes being
rotated for the new month because I want to make it as a archive for 1 year.
(let say restore for November before December start). Correct me if I wrong
for this understandings.
Or is it enough to me if I restore last full backup (26th) to represent for
the whole month? Or I have to choose last full backup and last incremental
backup instead?
There are a number of issues here that are hard to answer without more
information about your setup, volume of data and budget. These are
things that every backup administrator has to answer, and there is no
particular "right" answer.
One issue is how and when to do archives with Amanda. It's necessary to
realize that if you do an archive (say monthly), it is a snapshot in
time. There will be files that existed in between archives that are not
on the archives. I have arbitrarily decided to do archives at the end of
semesters, because I'm in an academic setting, and, typically, data,
accounts, etc. are likely to cycle before the beginning of the next
semester. During a semester, things are more fluid. I run a 6 week cycle
of tapes to give me fairly solid coverage in case I need to recover
something. I also tell faculty and staff that if they lose something,
the sooner they tell me, the better.
I have, in the past, just marked some tapes for no-reuse after examining
the reports to make sure every DLE was represented by a full on that set
of tapes. Then I add more tapes to fill out the cycle. I'm planning on a
slightly different tactic this Christmas. On a long weekend I'm going to
force all the DLEs to full, let it run its course, and then retire that
tape (or tapes) for archive by marking them no-reuse. I'll probably keep
end of semester archives for at least 2 years.
Once upon a time (mid 1990's), I ran daily backups (typical combination
of fulls and incrementals) and never re-used or discarded a tape, ever.
In that situation, I could go back to any day in the history of the
company and recover a file from any computer. But, I always had to buy
new tapes every month. With the explosion of data and storage
capacities, that is harder to do now. It takes much larger backup
capacity and higher budgets. If you have terrabytes of data, you're
talking serious money, and the bean counters are going to want to
balance the cost.
So, you filled(?) your 30 days of tapes, and now you want to put that
somewhere before you begin to fill another 30 days worth. How much data
is that? And where did you figure on storing it? If you are using tapes
in the first place, it seems fair to assume you don't have that much
excess disk space sitting around. If you want it all, then you could
just keep adding new tapes to your cycle. If you are not filling the
tapes, you could use a holding disk, configure Amanda so that it would
do both fulls and incrementals to the holding disk, and then only put a
tape in when the holding disk was approaching full. Amanda would
automatically flush it all out to tape on the next run (on the other
runs you'd get tape errors, but it would do the backups to the holding
disk). If your data capacity allowed you to put a tape in only once a
week, then 52 tapes would run your backups for a year.
You could also set up a separate archive configuration for Amanda,
schedule it to run once a month, and make sure the schedule doesn't
interfere with the normal daily runs. I prefer to have all the book
keeping and indexes in one configuration, but you may have other priorities.
There are a multitude of possible ways to configure backups with Amanda.
You should have enough information from running Amanda up to this point
to see how much data you have to backup and how well your tape capacity,
server, and network handle that. Then you put that together with your
budget and decide what is going to work for you for a long term strategy.
One point I would make, though, is that recovering from backup in order
to make an archive is probably not a good strategy. You would need a
large enough capacity disk to handle the recovery, and you would lose
the client machine information in the process, since the archive backup
would only know that it came from your recovery disk.
Hopefully, my comments along with Jon's will help you decide how to
configure your long term strategy.
---------------
Chris Hoogendyk
-
O__ ---- Systems Administrator
c/ /'_ --- Biology & Geology Departments
(*) \(*) -- 140 Morrill Science Center
~~~~~~~~~~ - University of Massachusetts, Amherst
<hoogendyk AT bio.umass DOT edu>
---------------
Erdös 4
|
|
|
