On Saturday 21 June 2003 18:14, Peter Kunst wrote:
>Hi,
>
>just migrating and testing how amanda goes on my setup, and i really
>like the level 0,1,... scheduling algorithm.
>However, after adding more and more partitions, i've tried to force
> a level 0 to all partitions and got on 50 percent of them errors
> like "planner: [dumps too big, xxx KB, but cannot incremental dump
> new disk]"
>
>How can i avoid these errors ?
>Or should i run "strategy nofull" ? But even then, i guess i
>might get the same errors when forcing a level 0 on all partitions.
>
>BTW, level 0 means more than 400GB here vs holding disk size of
>currently 20GB and an Overland LTO1 robot.
>
>Thanks for any hints (and yes, i've already searched on
> FAQ-O-Matic).
>
> Cheers,
> Peter
Aamnda is not designed to do the all-full situation at all. It *can*
be forced is there is sufficient tape capacity and you allow runtapes
to be whatever will hold it, in your case probably no less than 20-21
tapes.
IMO thats a bit unrealistic, particularly the resultant runtime which
will I'd guess, run well into the second day.
Amanda is designed to sprinkle the fulls around evenly during the
dumpcycle, and will, if left to her own scheduler, juggle things
around until amanda is using about the same amount of tape every
night. You give her a startup hand in that you only expose the
disklist entries for about 1/7th of the data per nightly run until
its all active, at which point you do nothing but keep the next runs
tapes in the robot, and read the emails to spot potential gotcha's.
This means that in the event of a disaster recovery, you'll need the
tapes from only the last 7 runs, or about 21 tapes if runtapes=3.
Your holding disk size should ideally be sufficient to hold one full
run, which in your case would be the size of the LT01 tape, I believe
you said its 20Gb (no, you didn't say, so thats probably an erronious
assumption on my part, scale the advice accordingly), times the
runtapes specified in your amanda.conf, plus whatever reserve level
you have spcified in amanda.conf for the holding disk.
With an LT01/20Gb tape and 400 Gb of data, I'd use an 80Gb drive or
partition with a 20% reserve and runtapes=3.
Based on the 20Gb guess, your *average* backup size of about 58Gb
however is going to need 3 of those tapes per night, so runtapes will
need to be set to 3 if dumpcycle is 7 days or 1 week. Thats based on
the 20Gb estimate for tapesize, so scale it to suit reality.
I'd setup a 120Gb disk as a holding disk since they are almost the
commodity drive today. It should also be on its own controller to
eliminate cable bus contentions which could slow things down
noticeably.
I'd also highly recommend a tapecycle (the number of tapes in the
pool) to be no less that 50 so that once up and running, you will
have two full backups of everything. You don't want to get into the
situation where the last full of something has been tapecycled into
reuse, and that backup session fails for whatever reason.
How many tapes does that LT01 robot carry at one time?
How big (in Gb) are they?
Whats the rated write speed of the drive?
Having that data will allow for more accurate prognostications.
Without it, the crystal balls antenna are obviously miss-adjusted. :)
--
Cheers, Gene
AMD K6-III@500mhz 320M
Athlon1600XP@1400mhz 512M
99.26% setiathome rank, not too shabby for a WV hillbilly
Yahoo.com attornies please note, additions to this message
by Gene Heskett are:
Copyright 2003 by Maurice Eugene Heskett, all rights reserved.
|