RE: Please help -- planner askfor: lev out of range -1..10: 10
You could at least temporarily go to 'runtapes 2' in amanda.conf to make
more room in each run. If you can afford more tapes, using two a day might
help, or using two tape drives with chg-multi might be good too.
Lengthening the tapecycle is going to take a while (about two tapecycles,
I'm guessing) to level out as you would expect. It sounds like you need more
tape now, which a second tape and/or a second drive could do.
Also, rather than regenerating the disklist every night, you might not
confuse amanda as much if you simply checked the directory structure every
night, and warned when a disklist entry is close to the limit. This would
allow you to plan splits better, and having your disk list entries change
less often is a good thing if you have to restore.
-Original Message-
From: Brashers, Bart -- MFG, Inc. [mailto:[EMAIL PROTECTED]]
Sent: Monday, April 22, 2002 11:52 AM
To: '[EMAIL PROTECTED]'
Subject: Please help -- planner askfor: lev out of range -1..10: 10
I didn't get any responses from my posting last Thursday, so
I'm guessing
I'm really up the creek on this one.
If anyone has any suggestions as to how to fix this problem,
please let me
know. I've now gone 4 days without a backup, and I'm afraid
I'm going to
have to delete everything and start from scratch!
Thanks,
Bart
After many months of operation (THANK YOU to all the developers!) last
night's amanda Daily run gave me this error:
FAILURE AND STRANGE DUMP SUMMARY:
planner: FATAL error [planner askfor: lev out of range -1..10: 10]
beowulf/usr/local RESULTS MISSING
beowulf/var/mail RESULTS MISSING
(and so on, for every element of the disklist)
The only debug files from this (9 PM) run were
/tmp/amanda.amtrmlog.*.debug
/tmp/amanda/amtrmidx.*.debug
but they showed nothing particularly interesting or helpful.
The latter
said
could not open index directory
/var/lib/amanda/Daily/index/db02/_root/
but that's because I recently added that machine-directory combination
(oops, forgot it) and amanda hasn't managed to schedule a
full dump, and
won't incrementally dump a new disk.
The relevant parts of my amanda.conf file are:
dumpcycle 3 weeks
runspercycle 15
tapecycle 17 tapes
define dumptype mydumptype {
program GNUTAR
compress none
} (plus more of the usual options)
I searched the archive, but could not find anything directly
relevant. This
lev out of range error happened to me once before, when I was first
learning and setting up amanda. That time, I just deleted
all the logs and
tapelists and everything, and started from scratch. I would
like to avoid
repeating that.
The only other (possibly) relevant detail is that I
constantly get errors
like:
FAILURE AND STRANGE DUMP SUMMARY:
beowulf/home/proj/mm5 lev 1 FAILED [dump larger than
tape, skipping
incremental]
beowulf/var/mail lev 1 FAILED [dumps way too big, must skip
incremental dumps]
(and so on, for ~30 or so directories)
I've written a script that runs (via crontab) every afternoon and
re-generates the disklist. It starts with a list of directories (/etc
/var/log /var/lib /usr/spool /root /home /data). For every
directory it
finds that contains more than 25Gb, it recursively breaks it
up and adds the
sub-directories to the disklist. This is because /home is
currently 153Gb,
and would never get dumped since it's larger than a tape.
I have about 200Gb to backup to a DLT8000 (40 Gb) tape drive.
I recently
went from a 2 to 3 week dumpcycle, hoping that the extra
space would get rid
of the dumps way too big error.
Please cc me with any hints as to how to clear this up, as I
only get daily
digests and I don't want to wait until tomorrow to fix this...
Bart
--
Bart Brashers MFG Inc.
Air Quality Meteorologist 19203 36th Ave W Suite 101
[EMAIL PROTECTED]Lynnwood WA 98036-5707
http://www.mfgenv.com 425.921.4000 Fax: 425.921.4040
RE: Please help -- planner askfor: lev out of range -1..10: 10
Thanks for your responses, Jon and Paul, I don't think going to 'runtapes 2' in amanda.conf is going to fix the lev out of range error, since that part happens long after my error. The email message starts out with FAILURE AND STRANGE DUMP SUMMARY: planner: FATAL error [planner askfor: lev out of range -1..10: 10] beowulf/usr/local RESULTS MISSING beowulf/var/mail RESULTS MISSING and there's nothing in /tmp/amanda/dump/20020419/ and the like. The planner fails, so nothing is ever gtar'd, so there's nothing to write to tape(s). Any ideas on how to clear the lev out of range error? Then I'll tackle the other errors stemming from having way to much data to backup compared to my tapesize. Bart, increasing the number of tapes should address this problem, because the planner is trying to figure out what level to run for each backup, and for unknown reasons is bouncing that backup down to a lower level than dump can handle. (Or one part of planner understands that 10 is the limit and another thinks that 9 is the limit.) Additional tape space will allow planner to go with a higher level for that (and other) backups. Paul
RE: Please help -- planner askfor: lev out of range -1..10: 10
--On Monday, April 22, 2002 11:42:46 -0600 Brashers, Bart -- MFG, Inc.
[EMAIL PROTECTED] wrote:
Hi Frank, thanks for your response!
To summarize what I think you've said, you are trying to back up
~200GB on a 40GB tape, and are dynamicly rewriting your disklist
entries into ~25GB chunks. You don't say how many tapes you have
in your runtapes and tapecycle, or if you have some kind of changer
configured (even if it is chg-manual).
The relevant parts of my amanda.conf file are:
dumpcycle 3 weeks
runspercycle 15
tapecycle 17 tapes
define dumptype mydumptype {
program GNUTAR
compress none
} (plus more of the usual options)
runtapes 1
tapedev /dev/nst0
I'm trying to avoid using chg-manual, and let amanda run on 1 tape per
night. I'd like to be able to use my tape drive for other non-amanda tasks
during the day ;-).
You could still use it during the day, as long as amdump or amcheck
weren't trying to use it. If you really don't want to set up a changer,
and if you have enough holding disk, you could lie to Amanda about
your tapetype and say it is 80GB, then planner will dump up to that
much. You will get a tape error in your daily report (since it will
hit EOT around 40GB), but you can then amflush it whenever it is
convenient.
FAILURE AND STRANGE DUMP SUMMARY:
beowulf/home/proj/mm5 lev 1 FAILED [dump larger than tape, skipping
incremental]
beowulf/var/mail lev 1 FAILED [dumps way too big, must skip
incremental dumps]
(and so on, for ~30 or so directories)
This says an incremental of /home/proj/mm5 is 40GB so it can't write
it to tape. The /var/mail error (dumps way too big) indicates that
the day's backup can't fit on the tapes allotted to it, so Amanda is
having to skip backups of those filesystems.
Ok, I'm confused. Each of these errors dump larger than tape and dumps
way too big refers not to the overall amanda run, BUT TO THAT PARTICULAR
DISK? That should be impossible. Each night _before_ amanda is run, my
ammakelist script would break up /home/proj/mm5 into sub-directories that
are each less than 25Gb. So it's impossible for the lev 1 to be larger than
25Gb, and it should fit on a tape. Not necessarily this tape, since the
amount to be written on any one night is generally larger than my
tapedrive...
[root]% du -h /var/spool/mail/
360k/var/spool/mail
There's no way the lev 1 on /var/spool/mail is larger than a tape. So these
errors dump larger than tape and dumps way too big must refer to the
total amount dumped to tape.
dumps way too big does refer to total backup size, but the other
warning, dump larger than tape refers to the individual filesystem.
Check the size of /home/proj/mm5. Perhaps its size changed between
the time your disklist was created and when amdump ran.
Increasing runtapes should help (although you may have to adjust
tapecycle and runspercycle unless you have extra tapes).
So my idea that I should add tapes (I recently went from 12 to 17) was
heading in the right direction. I have extra tapes in-house. Can I go to
dumpcycle 4 weeks
runspercycle 20
tapecycle 22 tapes
runtapes 1
or do I have to go to something like
dumpcycle 2 weeks
runspercycle 10
tapecycle 22 tapes
runtapes 2
Unless you can use compression to get more on a tape, it will take
quite a while to get everything backed up if you only use one tape
(5 or 6 days to do the first set of fulls without any incrementals).
I prefer having more than one full backup around, so I would prefer
the shorter dumpcycle. If your data doesn't change too much it might
only require two tapes for the first couple of weeks and then it
will level out and fit on one tape again. Definitely look into
compression, some data (especially text) compresses amazingly well.
If you are using hardware compression you may be able to increase
the size in your tapetype if you are not getting EOT errors.
much trimmed
Frank
--
Frank Smith[EMAIL PROTECTED]
Systems Administrator Voice: 512-374-4673
Hoover's Online Fax: 512-374-4501
