On Tue, Jun 27, 2006 at 10:09:05AM -0400, Brian Cuttler wrote:
> Jon,
> Stefan,
>
> yup, fixing the execute access in the directory path seems to have
> fixed the problem, I'm not only far ahead of where I was yesterday
> but I have a file in work waiting to flush (I do still need more
> work area as I'm currently dumping something direct to tape) but I
> can see that my dumpers took turns being busy.
>
> holding space : 25985401k ( 37.27%)
> dumper0 busy : 5:00:35 ( 42.43%)
> dumper1 busy : 0:25:27 ( 3.59%)
> dumper2 busy : 0:04:58 ( 0.70%)
> dumper3 busy : 0:53:23 ( 7.54%)
> dumper4 busy : 0:10:39 ( 1.50%)
> dumper5 busy : 0:30:10 ( 4.26%)
> dumper6 busy : 0:02:27 ( 0.35%)
> dumper7 busy : 0:11:12 ( 1.58%)
>
> They show really uneven usage but they where all utilized, a shame
> that I didn't preserve a complete amstatus output from one of the
> previous days.
>
> It maybe that I'm running an older version 2.4.2, but I haven't seen
> anything in the /tmp/amanda output, the amcheck nor the amstatus that
> would have made the fact that /amanda/work was unavailable explicite
> in any way -- which is not to say that we didn't do something really
> stupid that would have caused any utility mis-function.
>
> I currently have about 42Gig of free space with a DLE waiting to flush
> and another DLE writing directly to the tape. When the current one
> finishes the flush will begin, but will the last DLE, not yet started
> wait for the holding space or will it try to start a direct to tape ?
> Since its the last for the day it probably doesn't matter all that much
> from the "when will we be done" perspective, but from the wear and tear
> on the tape drive issue it does.
>
> Now to see about adding more work area... rebuild the device table
> (manually) to allow for more drives in the fiber array, then get a
> reboot scheduled (its a key system, so this will be fun to schedule).
>
> thank you,
>
> Brian
>
> On Mon, Jun 26, 2006 at 01:29:02PM -0400, Brian Cuttler wrote:
> > Jon,
> > Stefan,
> >
> > Suddenly...
> >
> > > cd /amanda/
> > > ls -l
> > total 46344
> > drw-rw-r-- 2 root root 8192 May 22 14:31 lost+found
> > drwxr-sr-x 7 root staff 512 Aug 22 2005 restore
> > -rw-r--r-- 1 root other 20497920 Jun 19 11:34 trimble-level0.tar
> > -rw-r--r-- 1 root other 3175936 Jun 19 11:34 trimble-level1.tar
> > drw-rw-r-- 2 amanda sys 512 May 5 02:27 work
> > drw-rw-r-- 2 amanda sys 512 Apr 12 09:22 workl
> > drw-rw-r-- 2 amanda sys 512 Apr 11 05:08 workr
> >
> > Now that isn't right... It used to be right... I don't know what
> > happened here.
Brian,
Paul B has submitted a patch to have amcheck check the execute
status of the holding disk area. So a future upgrade of your
server will probably have this problem caught.
BTW on your dumper usage, on my recently revised amanda system
I noted low percentage of multiple dumper usage. I made several
adjustments and it dropped my clock time by more than 50%. Among
them were upping total number of dumpers (was 4), dumpers per client
(was 1), and spreading my samba backups among all the unix/linux hosts.
I also reconfirmed my spindle numbers so that different parts of one
disk weren't backed up at the same time.
--
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)
|