Amanda-Users

Re: Weekly level 0 not completing

2009-07-24 07:59:18
Subject: Re: Weekly level 0 not completing
From: Jean-Louis Martineau <martineau AT zmanda DOT com>
To: Mark Fisher <mark.fisher AT nativ-systems DOT com>
Date: Fri, 24 Jul 2009 07:39:09 -0400
The tapetype definition and the complete amdump.X are needed to understand what amanda do wrong.
Maybe the estimate are bigger than the actual backup size, or ...

You can try to set 'maxdumpsize' to a value bigger than 1600GB.

Jean-Louis

Mark Fisher wrote:
Hi everyone,

We have been running amanda for a while now on our production kit
operating two similar configs, a daily and weekly schedule.

We run amanda v2.5.2p1 built from source, compress in software (a mix of
none, client and server depending on need & resources) and backup to a
HP MSL6000 library filled with LTO2 tapes (which of course are not using
hardware compression).  Slots 1-20 are used by the Dailies, 21-28 for
the Weeklies.

The Daily tapes stay in the library all the time, but we have two pools
of 8 Weekly tapes which are rotated to an offsite fire safe.  I use the
amadmin no-reuse command to 'flip' between pools (perhaps this is a bad
thing?).

The Daily backups work fine, but the Weekly setup has two problems.

Firstly, the dataset is now too big to fit on the 8 LTO2 tapes I can
make available for a weekly run.  We will be moving to either LTO3 or
more likely LTO4 in due course.

Secondly, amanda seems to estimate that the dataset to be backed up is
too large, complains that incrementals are not allowed and fails to
backup some of the smaller DLE's *.  I can understand that, but what I
don't understand is why it only _attempts_ to backup a given amount such
that only 7 of the configured 8 tapes are written to (and even then not
full up!) **.

[ .. ]

* seen in amstatus:
"0  planner: [dumps too big, 4356 KB, but no incremental estimate]"

** amanda.conf:
tapecycle 8 tapes
runtapes 8

amstatus:
taped           :  55   1344247m   1605218m ( 83.74%) ( 82.87%)
  tape 1        :  10    200324m    199580m ( 99.81%) weekly2_QKH895 (98
chunks)
  tape 2        :   3    197603m    428741m ( 98.46%) weekly2_QKH888
(100 chunks)
  tape 3        :   2    196970m    211500m ( 98.14%) weekly2_QKH889 (99
chunks)
  tape 4        :   1    198097m    198097m ( 98.70%) weekly2_QKH890
(100 chunks)
  tape 5        :   2    199398m    191011m ( 99.35%) weekly2_QKH891
(101 chunks)
  tape 6        :   2    200151m    200435m ( 99.72%) weekly2_QKH892
(102 chunks)
  tape 7        :  35    151701m    175853m ( 75.58%) weekly2_QKH893 (83
chunks)

[ .. ]

Is it possible that amanda makes these decisions on the uncompressed
sizes when the estimates first come in and not on the actual compressed
size as she runs?

Looking at the output of amadmin balance, it estimates that 1,936,414MB
is to be backed up, 1,352,664 compressed, which to me implies that the
~1,600,000MB available over the 8 LTO2's should fit if amanda tried?

I realise that I'm fighting a loosing battle in a sense, but
understanding this problem will help me understand amanda and ensure
that when we do invest in a new drive and tapes, that I budget
correctly.

Thanks,

--
  Mark Fisher



<Prev in Thread] Current Thread [Next in Thread>
  • Re: Weekly level 0 not completing, Jean-Louis Martineau <=