Amanda-Users

Re: newbie problems

2003-07-24 12:34:03
Subject: Re: newbie problems
From: Gene Heskett <gene.heskett AT verizon DOT net>
To: "Yogish" <yogish.gk AT ahsinc DOT com>, <amanda-users AT amanda DOT org>
Date: Thu, 24 Jul 2003 12:31:24 -0400
On Thursday 24 July 2003 10:56, Yogish wrote:
>Hi
>I was successfully able to perform a backup yesterday,thanks all of
> your help. I have a couple of doubts (trivial I'm sure). 1.How can
> I restore just a few files from a complete backup, for testing. 2.
> I want to do I complete backup,once a week and the incremental
> backup the remaining days,however when I try to backup data for the
> first time on tape, should I specify to amanda to do full backup or
> it will only do it. 3. Can I get a good documentation for Amanda?
> Please help if you can
>Regards
>Yogish

I'll let someone else discuss the recovery techniques, so I'll just 
address #2 above.  There are also a few links to some pretty good 
help that others will no doubt post.

Generally speaking, you will 'get along' with amanda a lot better if 
you let amanda do it her way.

The primary difference in how amanda works vs the rest of the world is 
that amanda can adjust the backup schedueling with a target of using 
the same amount of tape per run, prefereably nearly all of it.

That means that amanda will do some fulls, and some incrementals every 
run.

This is why you set such things as dumpcycle to be the amount of time 
that amanda has in order to do a full backup of everything in your 
disklist.  Set for 7 days, or 1 week, then amanda will attempt to get 
a full of everything in that 7 days.

Then you have to tell amanda how many runs she has in that 7 days 
because some people don't run it on the weekends, so if you don't, 
then 'runspercycle 5'

Then amanda needs to know how many tapes she can use at one time, this 
is 'runtapes' and is normally set to one unless you have both a 
changer so amanda can advance to the next tape, and enough tapes 
available.

Then finally, one needs to have at least enough tapes in the pool to 
allow at least (for good practices anyway) 2 full 
runspercycle*runtapes*2 tapes, prefereably even a few more.  This is 
called 'tapecycle'.

When first starting up amanda, only expose in the disklist, one tapes 
worth of entries per run, this because amanda must do a full before 
she can do an incremental, and by starting amanda up with a starter 
schedule, amanda will get into 'balance' with the tape useage per run 
a couple of dumpcycles faster.

Trying to make amanda do a full on friday night, and incrementals the 
rest of the week can be done, but not only wastes tapes by doing 
wildly differing sizes of backups depending on the run, and labeling 
the tapes for the day of the week you will find will be an exersize 
in futility.  The tape useage will be out of step with the tape 
labels in 2 weeks or less, for any one of a hundred reasons.

I'm just a home user, running dirt cheap DDS2 tapes in a 4 tape 
changer so I have:

dumpcycle 7 days
runspercycle 7
runtapes 1
tapecycle 28

This is adequate to cover around 33Gb of real data on 2 machines, with 
40 some entries in the disklist, and using 90 or more percent of a 4 
Gb DDS2 tape (hardware compression is off, software is on for 
selected disklist entries) each night.

You will probably have more data, probably much more, so the tape 
drive and tape capacity would be scaled up accordingly.  Since its an 
axiom of the business that tape speeds go up pretty much in step with 
the capacity, then the actual times to do these backups remains 
fairly consistent.  With my old slow DDS2 setup, its about a 2.5 hour 
run each night.

I mentioned hardware compression being turned off, and recommend that 
it be off for several reasons.

1. with it on, amanda's count of bytes fed to the drive is meaningless 
because the drive is modifying that invisibly to amanda.  This means 
that to prevent hitting the EOT of the tape, you have to reduce the 
size of the tape in its tapetype by some fudge factor you can only 
guess.

2. With hardware off, and software being used on those DLE's that can 
be compressed (a directory full of .bz2;s, .gz's, and rpm's will not 
further compress and may even grow some) can be compressed better 
than the hardware can do, sometimes much better, so you use the 
tapetype sizeing for the native tape size.  Amanda then knows how big 
the tape is, and can stuff it to the brim.  The tradeoff of course is 
the cpu horsepower to do the compression.  However this becomes a 
never mind when you can offload the compression to the client machine 
because then several compressors can be running in parallel, and the 
network bandwidth is reduced by the resultant smaller files being 
shipped back to the server.

-- 
Cheers, Gene
AMD K6-III@500mhz 320M
Athlon1600XP@1400mhz  512M
99.27% setiathome rank, not too shabby for a WV hillbilly
Yahoo.com attornies please note, additions to this message
by Gene Heskett are:
Copyright 2003 by Maurice Eugene Heskett, all rights reserved.


<Prev in Thread] Current Thread [Next in Thread>