Amanda-Users

Re: What will i do if a backup is not successfull?

2005-06-04 02:18:27
Subject: Re: What will i do if a backup is not successfull?
From: Jon LaBadie <jon AT jgcomp DOT com>
To: Mailing List Amanda User <amanda-users AT amanda DOT org>
Date: Sat, 4 Jun 2005 02:06:37 -0400
On Sat, Jun 04, 2005 at 10:36:47AM +0800, Ryan Pagquil wrote:
> Paul Bijnens wrote:
> 
> >Ryan Pagquil wrote:
> >
> >>>2005/6/3, Jon LaBadie <jon AT jgcomp DOT com>:
> >>>
> >>>>man amdump shows:
> >>>>
> >>>> SYNOPSIS
> >>>>    amdump config [ host [ disk ]* ]*
> >>>>
> >
> >>Hi guys,
> >>      The scenario is this, one night one of the servers to be backed 
> >>up hanged and it will not be dump by amanda, but all of the other 
> >>servers are successfully dump. In the morning i want that server to 
> >>be backed up so that all of my backups are synchronize, meaning no 
> >>servers are left behind the backup schedule. Can I do this? and how 
> >>can I deal with this kind of scenario?
> >
> >
> >Execute by hand:
> >
> >    amdump Config the.fixed.server
> >
> >You may insert a tape, if you like.  But setting "reserve" less than
> >100%, and a large enough holdingdisk, you may dump to holdingdisk only,
> >and with the "autoflush" parameter in amanda, write it to tape, the
> >next run together with the rest.
> 
> Hi Paul,
>       I may use a new  tape or one tape on rotation (eg. the next tape 
> on rotation) right? Once i've done this, will amanda also recognize what 
> tape i put the dumps from that single server? But when i use the 
> holdingdisk to store the dumps (in your suggestion), the dump(for that 
> failed server)  will be flushed on to tape on the next run of amanda 
> including the regular dump for that server scheduled for that run?

I think a difficulty is that there are several possibilities.

1. Don't do anything but run your normal amdump schedule.
   Amanda knows the status of the failed server and will bring
   it up to date the next amdump run.

2. Run an extra amdump as soon as you have the server online
   and send it to tape.  For this you should have a tape in
   the drive while amdump is started.  It can be the next
   tape from your normal rotation or an extra "new" tape.
   The next time around in the tape rotation, the new tape
   will be expected in the same order.  But if you like, you
   could mark it for noreuse so that does not happen.

3. Run an extra amdump like above but without a tape in the
   drive.  This will cause the dump to collect on the holding
   disk for later taping.  The later taping can be done with
   amflush, or if configured appropriately it will happen
   along with the next amdump.  For this to work well you need
   a holding disk large enough to hold the extra dump plus the
   normal load of the later amdump; the reserve parameter must
   be set to allow the holding disk to be used for level 0's
   if needed (say 10 - 50%); and the autoflush parameter must
   be set to yes.

If you are backing up many clients you may find it a ?common?
occurence that one or more clients, or some DLEs are not
available at the time amdump runs.  For extremely critical
data you may want to do the extra amflush.  If so, make sure
it will have completed by the time your regular amdump does
run.  Alternatively consider waiting for the next regular
amdump.  The missed DLEs won't upset amanda.  It will adjust
and do the "right" thing the next run.

-- 
Jon H. LaBadie                  jon AT jgcomp DOT com
 JG Computing
 4455 Province Line Road        (609) 252-0159
 Princeton, NJ  08540-4322      (609) 683-7220 (fax)