Re: Dramatic reduction in backup time
2006-08-03 11:59:49
On Thu, Aug 03, 2006 at 08:25:43AM -0700, Joe Donner (sent by Nabble.com) wrote:
>
> Well, I may have "lied" a bit when saying that nothing had changed...
>
> What did change was that I moved the server from temporarily sitting on the
> floor to where it will now stay full-time. So it was a shutdown, unplug the
> external tape drive and other devices, and then plug them all back in.
> Also, one of the little screws which connects the SCSI cable to the back of
> the server has long ago broken off and gone missing. So it's only attached
> by one screw.
>
> Well, do you agree with me that it is still backing up the same amount of
> data and all the DLEs? It certainly looks that way from the mail reports.
> I'm a bit worried because it's looking "too good to be true".
>
> I also don't see how it can dump 130-odd gigabytes of data within a bit more
> than 3 hours??
>
> Thanks for your help on this.
>
>
> Jon LaBadie wrote:
> >
> > On Thu, Aug 03, 2006 at 02:46:06AM -0700, Joe Donner (sent by Nabble.com)
> > wrote:
> >>
> >> That was definitely not the case. The holding disk has been empty all
> >> this
> >> time, except on 31 July. And after I ran amflush it was empty again.
> >>
> >> What I don't understand is the increase in dump time and tape time? They
> >> both more than doubled in speed! How is that possible?
> >>
> >>
> >> Olivier Nicole wrote:
> >> >
> >> >> I honestly haven't changed anything. The only thing that happened was
> >> >> that
> >> >> I had to run amflush on July 31, because for some reason the
> >> >> amanda-related
> >> >> services on the backup server seemed to have hung (but that's an
> >> entirely
> >> >> different issue).
> >> >
> >> > That si just a wilkd guess, but how about the amflush has freed space
> >> > on your holding disk that had been wasted for ages?
> >> >
> >
> > Holding disk issues were going to be my guess also.
> >
> > When the run time (wall clock) and tape time (taper
> > is active) match as the first two:
> >
> > Run Time (hrs:min) 9:43
> > Dump Time (hrs:min) 6:21
> > Tape Time (hrs:min) 9:23
> >
> > Run Time (hrs:min) 9:43
> > Dump Time (hrs:min) 6:17
> > Tape Time (hrs:min) 9:23
> >
> > it "generally" means your dumps are going straight to tape.
> > Contrast that with the last two where time for taping the
> > same amount of data is greatly lower than total run time.
> >
> > Run Time (hrs:min) 4:03
> > Dump Time (hrs:min) 3:12
> > Tape Time (hrs:min) 1:17
> >
> > Run Time (hrs:min) 4:27
> > Dump Time (hrs:min) 3:37
> > Tape Time (hrs:min) 1:16
> >
> > However, if the dumps were going straight to tape then I
> > would expect the dump time (cumulative time for dumping of
> > each DLE - thus can be > run time if DLEs dump in parallel)
> > to match tape time.
> >
> > So it would appear the dumps were going to the holding disk
> > but were severely constrained by slow tapeing. I wonder
> > about the halving of dump time also. There may still have
> > been a problem with the disk system.
> >
> > Did anyone change any scsi cables or terminators.
> > Or maybe the just "jiggled" things back there doing other work?
> >
> >
Are you using that SDLT drive for which you reported a tapetype
back in early June? If so, you have had a problem since then.
In June your tapetype showed 2200 k/s.
Similarly, your recent slow dumps had a
tapeing speed of about 2500 and 1800 k/s.
Your more recent fast dumps were taped at about 13000 k/s.
The sdlt is rated at 16000 k/s.
I'd say you had tape system problems
fixed by physical movement or reboot.
--
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)
|
|
|