Amanda-Users

Re: FW: dump level problems

2007-09-11 14:44:16
Subject: Re: FW: dump level problems
From: Jon LaBadie <jon AT jgcomp DOT com>
To: amanda-users AT amanda DOT org
Date: Tue, 11 Sep 2007 14:43:27 -0400
On Tue, Sep 11, 2007 at 05:14:24PM +0200, Mark Scheufele wrote:
> Hi,
> 
> I am having the problem that our amanda installation doesn't increment
> the dump levels for the filesystems defined in the DLE.
> 
> This is the backup status mail from Friday 7th of September:
> 
> NOTES:
>   planner: Last full dump of nessie.diasemi.com:/it on tape Weekly-03
> overwritten in 3 runs.
>   planner: Full dump of nessie.diasemi.com:/users promoted from 7 days
> ahead.
>   planner: Full dump of nessie.diasemi.com:/projects promoted from 6
> days ahead.
>   planner: Full dump of nessie.diasemi.com:/design promoted from 7 days
> ahead.
>   taper: tape Weekly-07 kb 5666848 fm 4 [OK]
> 
> 
> DUMP SUMMARY:
>                                        DUMPER STATS               TAPER
> STATS 
> HOSTNAME     DISK        L ORIG-KB  OUT-KB  COMP%  MMM:SS   KB/s MMM:SS
> KB/s
> -------------------------- -------------------------------------
> -------------
> nessie.diase /design     0    3296    3296    --     0:08  388.6   0:09
> 373.8
> nessie.diase /it         1  136610  136640    --    12:56  176.1  12:56
> 176.1
> nessie.diase /projects   0 4298144 4298144    --    31:56 2243.2  31:56
> 2243.1
> nessie.diase /users      0 1228640 1228640    --     8:47 2330.2   8:48
> 2328.2
> 
> NOTES:
>   planner: Last full dump of nessie.diasemi.com:/it on tape Weekly-03
> overwritten in 2 runs.
>   planner: Full dump of nessie.diasemi.com:/users promoted from 4 days
> ahead.
>   planner: Full dump of nessie.diasemi.com:/projects promoted from 4
> days ahead.
>   planner: Full dump of nessie.diasemi.com:/design promoted from 4 days
> ahead.
>   taper: tape Weekly-01 kb 5666464 fm 4 [OK]
> 
> This the backup status mail from the 10th of September.
> 
> DUMP SUMMARY:
>                                        DUMPER STATS               TAPER
> STATS 
> HOSTNAME     DISK        L ORIG-KB  OUT-KB  COMP%  MMM:SS   KB/s MMM:SS
> KB/s
> -------------------------- -------------------------------------
> -------------
> nessie.diase /design     0    3296    3296    --     0:09  374.7   0:09
> 363.5
> nessie.diase /it         1  137490  137504    --    21:28  106.7  21:29
> 106.7
> nessie.diase /projects   0 4297664 4297664    --    38:12 1875.0  38:12
> 1874.7
> nessie.diase /users      0 1227872 1227872    --    10:26 1960.1  10:27
> 1959.1
> 
> (brought to you by Amanda version 2.5.2p1)
> 
> I am now wondering why the levels aren't incremented? I have configured
> a dumpcycle of 7 days and a runcycle of 5 days. I am using ufsdump as
> the actual dump program. The contents of /etc/dumpdates is:
> 
> [root@nessie:/]#cat /etc/dumpdates 
> /dev/zvol/rdsk/edpool/users      0 Mon Sep 10 20:38:40 2007
> /dev/zvol/rdsk/edpool/projects   0 Mon Sep 10 20:48:58 2007
> /dev/zvol/rdsk/edpool/design     0 Mon Sep 10 21:27:11 2007
> 
> /etc/amdates looks as follows:
> 
> /design 0 1189194562
> /design 1 1188901535
> /it 0 1188890901
> /it 1 1189451823
> /projects 0 1189192646
> /users 0 1189192125
> /users 1 1188901421
> 
> Any hints are very appreciated.
> 

It looks to me like you just switched from gnutar to ufsdump 
as your backup program.  /etc/amandates (I presume you mis-
typed the file name) is used by amanda for gnutar while
/etc/dumpdates is used by ufsdump.

A switch may have caused a "forced" level 0 using the new
program.  I note the "it" still recorded in /etc/amandates
and, guessing again, it did not use ufsdump.

I note the "/dev/zvol" for three of the DLE.  Are you using
zfs tools to create slices/partitions and then formating the
space with a ufs file system??  I ask because if they are
really zfs file systems, I don't think ufsdump will work
properly.

I presume all these four DLE are mounted file systems.  If
not, and they are directory trees, be aware that ufsdump
can not do incrementals on anything except full file systems.

You say "a runcycle of 5 days", but I'm sure you mean
"runspercycle".  Be aware that parameter does not take a
unit.  It is 5, not 5 days.

If you only have 4 DLE and a longer dumpcycle and runspercycle,
you should expect to have lots of promotion.  Amanda will be
trying to balance the amount of data taped each run.  The only
way to achieve this balance is by moving around the level 0's.
If this is close to your final configuration, you may wish to
consider an administrator-determined schedule rather than
letting amanda do the scheduling.  Review in past discussions
about setting your dumptypes to incremental only and forcing
the level 0's on days you choose.

-- 
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)

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