On Fri, Apr 29, 2005 at 12:11:37AM -0400, Jon LaBadie enlightened us:
> > My current setup is an LTO1 tape drive that I wouldn't even come close to
> > filling on a nightly basis. I get right around the 100GB tape size in about
> > 6 or 7 days, so I leave previous tapes in for about a week, letting files
> > build up on the holding disk, and then load new tapes once a week letting
> > autoflush take care of it, or if I've forgotten and gone over 100GB, I'll
> > run an amflush by hand.
> >
> > I noticed today that the largestfit taperalgo wasn't doing what I expected
> > it to. Below are the relevent snippet from the mail report, and the contents
> > of my dump directory:
> >
> > FAILURE AND STRANGE DUMP SUMMARY:
> > bing.math. /var lev 0 FAILED [out of tape]
> >
> > NOTES:
> > taper: tape Dailies09 kb 104234048 fm 114 writing file: No space left
> > on device
> >
> > So we wrote about 1GB of data before running out of tape.
> >
>
> I don't see where you are getting 1GB from.
>
Funny, me either. I think I read that too quickly and saw it as about half
of the 2.4GB dump it failed on.
> > [root@bing root]# ls -l /dumps/Dailies/20050428/
> > total 6773656
> > -rw------- 1 amanda disk 651761587 Apr 28 01:12
> > bing.math.ohiou.edu._export.3
> > -rw------- 1 amanda disk 507741725 Apr 28 01:24
> > bing.math.ohiou.edu._export_profiles.5
> > -rw------- 1 amanda disk 2457299460 Apr 28 01:30
> > bing.math.ohiou.edu._var.0
> > -rw------- 1 amanda disk 474850788 Apr 28 01:07
> > eric.math.ohiou.edu._var.0
> > -rw------- 1 amanda disk 555816686 Apr 28 01:05
> > euclid.math.ohiou.edu._.1
> > -rw------- 1 amanda disk 1700676298 Apr 28 00:56
> > turing.math.ohiou.edu._var.0
> > -rw------- 1 amanda disk 581250639 Apr 28 01:08
> > young.math.ohiou.edu._home.0
> >
> > The dump it failed on was bing /var, which is the largest image here,
> > however I would have expected amanda to know there was only about 1GB of
> > space left on the device, so it should have done something smaller. If I
> > were amanda, I would have said that bing /var and turing /var were too big
> > at 2.4 and 1.7GB respectively, and tried the 650MB bing /export.
> >
> > Obviously I am not amanda, and that is not what "she" tried to do, so can
> > anyone enlighten me? What is the difference between largest and largestfit
> > taperalgos?
>
> IIRC LTO is a 100GB native tape. I assume that is about what
> you have your tapetype set to. At least, I'm guessing you
> have it set to less than 104.234048 GB.
>
> If that is the case, your tapetype claims the capacity is less than
> the amount amanda has already successfully written to the tape,
> what is the meaning of "remaining space" on an appearently already
> filled tape that has not reached the end. I'd say no largestfit
> algorithm would work under that scenario.
Yes, as Paul Bijnens pointed out in a private mail, I had already gone over
what amanda knew about. In that case, largestfit falls back to the largest
dump available.
I'm waiting to hear back from him, but Paul said he modified his
installation to fall back to the smallest dump available in that case, in
the hope that it might fit on the tape. I think that makes the most sense to
do.
I've enabled hardware compression on my drive after having it off for the
last 2 years since it doesn't hurt with LTO, and if anything it might
squeeze a little more data on there if I turn client software compression
off at some point (one machine in particular is starting to get *extremely*
slow and is holding up the rest of the group). Since that is the case, there
will likely always be a little more space on the tape than amanda thinks
since some of the data may have been compressed.
As for me, I'll try to read those numbers a little more carefully before
posting next time.
Thanks,
Matt
--
Matt Hyclak
Department of Mathematics
Department of Social Work
Ohio University
(740) 593-1263
pgpLlpj4iodAt.pgp
Description: PGP signature
|
