Amanda-Users

archival runs - very large holding disk strategy

2003-04-28 20:26:08
Subject: archival runs - very large holding disk strategy
From: Greg Troxel <gdt AT ir.bbn DOT com>
To: amanda-users AT amanda DOT org
Date: 28 Apr 2003 20:22:05 -0400
Periodically people ask how to make tapes for offsite or archival
backups, and there are several approaches mentioned.  I just started a
new (to me anyway) method, and thought I'd mention it here.

The key insight is that running dumps and having the tape drive
available with the right tape don't have to be coupled at all if you
have > tapesize of holding disk.

I have 38 GB on 53 filesystems on 11 machines being backed up to a
DDS3 drive.  I run 5 tapes per week on the mornings following
weekdays, with a dumpcycle of 7 days.  On Saturday night, amverify
checks Friday night's dump.

Also on Saturday night (Sunday morning really), delayed by about 10
minutes to allow the amverify to grab the tape drive, I run amdump on
the archive configuration.  The archive configuration is the same,
except it has 'skip-incr', 'record no', 'reserve 0', a
dumpcycle/runspercycle of 14/2, and a different holding disk
directory.  This may get behind, but I'll fill the tape every week.

The archive dump fails to find a tape, and dumps to holding disk, of
which I have 40-50 GB available.  Then, on Monday morning, I put in an
archive tape and amflush the archive configuration, which finishes
during the day so I can stick in the daily tape.  In this way, I get
one run per week of full dumps onto a different set of tapes from the
normal rotation.

I realize now that amverify does not hold the tape drive open, so the
strategy of starting the archive amdump later is wrong, so I just
changed the crontab to do amdump at 0005 and amverify at 00015.

        Greg Troxel <gdt AT ir.bbn DOT com>

<Prev in Thread] Current Thread [Next in Thread>
  • archival runs - very large holding disk strategy, Greg Troxel <=