|
Home > Archive > Unix administration > July 2005 > How to limit a amanda backup level?
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
How to limit a amanda backup level?
|
|
| Dragan Cvetkovic 2005-07-11, 5:53 pm |
| Hi,
we are using amanda to backup a disk with a lot of small changes (Solaris,
LTO2 tapes in a Seagate 7 tape changer), and amanda is incrementing backup
level quite often, so that it currently does level 8 backups! Quite a few
ones to restore:
bash-2.03# /usr/local/sbin/amadmin v880 info myclient /mydisk
Current info for myclient /mydisk:
Stats: dump rates (kps), Full: 1979.0, -1.0, -1.0
Incremental: 1209.0, 1544.0, 1492.0
compressed size, Full: 19.6%,-100.0%,-100.0%
Incremental: 27.4%, 21.3%, 21.8%
Dumps: lev datestmp tape file origK compK secs
0 20050606 tape_01 2 14056191 2748884 1389
1 20050608 tape_03 2 7588863 1213215 700
2 20050614 tape_02 2 328831 53616 37
3 20050616 tape_04 2 6964735 1079838 624
4 20050620 tape_01 2 217855 33663 25
5 20050622 tape_03 2 7724031 1155358 694
6 20050629 tape_03 2 7113631 1096765 643
7 20050706 tape_03 2 7040447 1088322 629
8 20050711 tape_01 2 52863 14509 12
Bye, Dragan
--
Dragan Cvetkovic,
To be or not to be is true. G. Boole No it isn't. L. E. J. Brouwer
!!! Sender/From address is bogus. Use reply-to one !!!
| |
| Mike Delaney 2005-07-11, 8:48 pm |
| On Mon, 11 Jul 2005 16:58:25 -0400 in <lmoe99xdou.fsf@privacy.net>,
Dragan Cvetkovic said something similar to:
: Hi,
:
: we are using amanda to backup a disk with a lot of small changes (Solaris,
: LTO2 tapes in a Seagate 7 tape changer), and amanda is incrementing backup
: level quite often, so that it currently does level 8 backups! Quite a few
: ones to restore:
That's alot of changes - ~ 7G on level 1, 3, 5, 6, and 7 dumps of a filesytem
whose previous level 0 was ~ 14GB.
And what's up with amadmin reporting multiple different dumps occupying the
same filemark on the same tape (or have you been editing the output)?
At any rate, the parameters that control promotion are bumpsize, bumpmult, and
bumpdays - see amanda(8). There's also a per-dumptype option, maxpromoteday,
but the documentation for it is a little terse.
If I were in your place, I'd consider reducing the dumpcycle - from the
amadmin output you posted, it appears to be well over a month. Is there
a particular reason you're running that long between level 0s?
| |
| Dragan Cvetkovic 2005-07-12, 7:56 am |
| Mike Delaney <mdelan@computer.org> writes:
> On Mon, 11 Jul 2005 16:58:25 -0400 in <lmoe99xdou.fsf@privacy.net>,
> Dragan Cvetkovic said something similar to:
> : Hi,
> :
> : we are using amanda to backup a disk with a lot of small changes (Solaris,
> : LTO2 tapes in a Seagate 7 tape changer), and amanda is incrementing backup
> : level quite often, so that it currently does level 8 backups! Quite a few
> : ones to restore:
>
> That's alot of changes - ~ 7G on level 1, 3, 5, 6, and 7 dumps of a filesytem
> whose previous level 0 was ~ 14GB.
Yes, there are a lot of scratch files being created and deleted dayly and I
don't want to use tar instead of ufsdump because of performance
reasons. Must see if there is enough disk space so that I can move that
particular directory to a separate slice and simply not back it up.
> And what's up with amadmin reporting multiple different dumps occupying the
> same filemark on the same tape (or have you been editing the output)?
That confused me as well. We only have 5 tapes here and it seems that
amanda doesn't append to a tape as it should (and as it does on other
systems). Will have to investigate that further.
> At any rate, the parameters that control promotion are bumpsize, bumpmult, and
> bumpdays - see amanda(8). There's also a per-dumptype option, maxpromoteday,
> but the documentation for it is a little terse.
I have tried to play a bit with it, increased bumpsize from 10MB to 100MB
and bumpmult from 2 to 4. Let's see if that helps.
> If I were in your place, I'd consider reducing the dumpcycle - from the
> amadmin output you posted, it appears to be well over a month. Is there
> a particular reason you're running that long between level 0s?
Well, I would like to have a retention period of 8 weeks and thought that
dumpcycle is the one to use for it. Is there any other parameter for
specifying retention?
Thanks for your help so far.
Bye, Dragan
--
Dragan Cvetkovic,
To be or not to be is true. G. Boole No it isn't. L. E. J. Brouwer
!!! Sender/From address is bogus. Use reply-to one !!!
| |
| Mike Delaney 2005-07-12, 5:57 pm |
| On Tue, 12 Jul 2005 08:50:46 -0400 in <lm4qb0xk61.fsf@privacy.net>,
Dragan Cvetkovic said something similar to:
: Mike Delaney <mdelan@computer.org> writes:
:
: > And what's up with amadmin reporting multiple different dumps occupying the
: > same filemark on the same tape (or have you been editing the output)?
:
: That confused me as well. We only have 5 tapes here and it seems that
: amanda doesn't append to a tape as it should (and as it does on other
: systems). Will have to investigate that further.
Amanda never appends to tapes. Every run starts at the beginning of a
new tape. You've got a bunch of incrementals with no baseline level 0
because you've configured Amanda to overwrite the tapes more frequently
than level 0 dumps are performed. Amanda should have complained about that
in the report the day the level 0 was overwritten.
: > If I were in your place, I'd consider reducing the dumpcycle - from the
: > amadmin output you posted, it appears to be well over a month. Is there
: > a particular reason you're running that long between level 0s?
:
: Well, I would like to have a retention period of 8 weeks and thought that
: dumpcycle is the one to use for it. Is there any other parameter for
: specifying retention?
The dumpcycle and runspercycle parameters determine the frequency of level 0
dumps.
dumpcycle = 7 days
runspercycle = 7
This tells Amanda that a level 0 dump is to be performed for each disklist
entry (DLE) at least once a week, and that amdump is going to be run 7
times during that week (i.e. every day).
The tapecycle and runtapes parameters are what control the retention period.
Amanda will not overwrite a tape until she has gone through at least $tapecycle
tapes. For one config I maintain,
tapecycle = 30
runtapes = 2
With the example above, the minimum bound on the retention period is:
(30 tapes) / [ (1 dump/day) * (2 tapes/dump) ] = 15 days
If Amanda doesn't need 2 tapes for each run, the retention in this example
could be as high as 30 days. (Or higher, if the number of tapes in the
rotation is larger than $tapecycle.)
If you're doing dumps every night and want 8 weeks retention, then you
need at least 56 tapes to do that.
:
: Thanks for your help so far.
:
: Bye, Dragan
:
| |
| Dragan Cvetkovic 2005-07-12, 5:57 pm |
| Mike Delaney <mdelan@computer.org> writes:
> Amanda never appends to tapes. Every run starts at the beginning of a
> new tape. You've got a bunch of incrementals with no baseline level 0
> because you've configured Amanda to overwrite the tapes more frequently
> than level 0 dumps are performed. Amanda should have complained about that
> in the report the day the level 0 was overwritten.
Ah, I see.
[snip]
> The dumpcycle and runspercycle parameters determine the frequency of level 0
> dumps.
>
> dumpcycle = 7 days
> runspercycle = 7
>
> This tells Amanda that a level 0 dump is to be performed for each disklist
> entry (DLE) at least once a week, and that amdump is going to be run 7
> times during that week (i.e. every day).
>
> The tapecycle and runtapes parameters are what control the retention period.
> Amanda will not overwrite a tape until she has gone through at least $tapecycle
> tapes.
[snip]
> If you're doing dumps every night and want 8 weeks retention, then you
> need at least 56 tapes to do that.
OK, that now makes sense. What confused me quite a bit was that we have a
configuration here with
dumpcycle 4 weeks # the number of days in the normal dump cycle
runspercycle 20 # the number of amdump runs in dumpcycle days
# (4 weeks * 5 amdump runs per week -- just weekdays)
tapecycle 10 tapes # the number of tapes in rotation
and we still get several level 0 dumps (backup is every workday):
% /usr/local/sbin/amadmin tapeconf find titan /export/home/titan
Scanning /amanda/work...
date host disk lv tape or file file status
2004-07-02 titan /export/home/titan 2 tape99 38 OK
2005-06-29 titan /export/home/titan 1 tape06 49 OK
2005-06-30 titan /export/home/titan 2 tape07 46 OK
2005-07-01 titan /export/home/titan 2 tape08 55 OK
2005-07-02 titan /export/home/titan 2 tape05 25 OK
2005-07-05 titan /export/home/titan 0 tape09 65 OK
2005-07-06 titan /export/home/titan 1 tape00 54 OK
2005-07-07 titan /export/home/titan 2 tape01 44 OK
2005-07-08 titan /export/home/titan 2 tape02 53 OK
2005-07-09 titan /export/home/titan 2 tape03 50 OK
2005-07-12 titan /export/home/titan 0 tape04 65 OK
Bye, Dragan
--
Dragan Cvetkovic,
To be or not to be is true. G. Boole No it isn't. L. E. J. Brouwer
!!! Sender/From address is bogus. Use reply-to one !!!
| |
| Mike Delaney 2005-07-12, 5:57 pm |
| On Tue, 12 Jul 2005 15:05:52 -0400 in <lmirzfx2sv.fsf@privacy.net>,
Dragan Cvetkovic said something similar to:
: Mike Delaney <mdelan@computer.org> writes:
:
: > If you're doing dumps every night and want 8 weeks retention, then you
: > need at least 56 tapes to do that.
:
: OK, that now makes sense. What confused me quite a bit was that we have a
: configuration here with
:
: dumpcycle 4 weeks # the number of days in the normal dump cycle
: runspercycle 20 # the number of amdump runs in dumpcycle days
: # (4 weeks * 5 amdump runs per week -- just weekdays)
: tapecycle 10 tapes # the number of tapes in rotation
You should be getting warnings from the planner every run with that:
"tapecycle (10) <= runspercycle(20)"
The tapecycle parameter should always be set to a larger value than
runspercycle.
:
: and we still get several level 0 dumps (backup is every workday):
:
Amanda's scheduling algorithm might be increasing the frequency of the
level 0s to make up for the short tapecycle. Dumpcycle tells Amanda
the minimum frequency with which a level 0 must be done. She's quite
free to do them more frequently if the tape capacity is there.
: % /usr/local/sbin/amadmin tapeconf find titan /export/home/titan
: Scanning /amanda/work...
[snip]
This isn't the same filesystem as the origional post is it? If you've
got a lower rate of change here than on the filesystem from your OP,
that would account for why this DLE is working with the broken config
but the other isn't.
| |
| Dragan Cvetkovic 2005-07-13, 7:55 am |
| Mike Delaney <mdelan@computer.org> writes:
> On Tue, 12 Jul 2005 15:05:52 -0400 in <lmirzfx2sv.fsf@privacy.net>,
> Dragan Cvetkovic said something similar to:
> : Mike Delaney <mdelan@computer.org> writes:
> :
> : > If you're doing dumps every night and want 8 weeks retention, then you
> : > need at least 56 tapes to do that.
> :
> : OK, that now makes sense. What confused me quite a bit was that we have a
> : configuration here with
> :
> : dumpcycle 4 weeks # the number of days in the normal dump cycle
> : runspercycle 20 # the number of amdump runs in dumpcycle days
> : # (4 weeks * 5 amdump runs per week -- just weekdays)
> : tapecycle 10 tapes # the number of tapes in rotation
>
> You should be getting warnings from the planner every run with that:
> "tapecycle (10) <= runspercycle(20)"
>
> The tapecycle parameter should always be set to a larger value than
> runspercycle.
Yes, I understand that now.
>
> :
> : and we still get several level 0 dumps (backup is every workday):
> :
>
> Amanda's scheduling algorithm might be increasing the frequency of the
> level 0s to make up for the short tapecycle. Dumpcycle tells Amanda
> the minimum frequency with which a level 0 must be done. She's quite
> free to do them more frequently if the tape capacity is there.
I guess there are too many tunables here :-)
The fact that I have started using amanda and veritas netbackup (which does
append to a tape) more or less at the same time, didn't really help. I
probably mixed a concept or two.
I think that I'll need to modify a few scripts...
>
> : % /usr/local/sbin/amadmin tapeconf find titan /export/home/titan
> : Scanning /amanda/work...
>
> [snip]
>
> This isn't the same filesystem as the origional post is it?
No, we have something like 4 or 5 different amanda based backup systems.
> If you've
> got a lower rate of change here than on the filesystem from your OP,
> that would account for why this DLE is working with the broken config
> but the other isn't.
Mike, thanks a lot for the clarification, I appreciate it.
Bye, Dragan
--
Dragan Cvetkovic,
To be or not to be is true. G. Boole No it isn't. L. E. J. Brouwer
!!! Sender/From address is bogus. Use reply-to one !!!
|
|
|
|
|