>
> Several things.
>
> 1. I suspect you are within your first 1 or 2 dumpcycles.
>    Balance will take several dumpcycles to achieve.

You are correct sir!. I'm in the 2 dumpcycle. I figured this might play into
the appearant imbalance.

> 2. +/-5% is a target, seldom achived except at large sites with
>    lots of hosts and file systems being backed up.

Understood. I was just a little shocked by the ~260%.

> 3. I think the output reports file systems assuming the next
>    level 0 will be exactly 1 dumpcycle since the last one.
>    Yet when the actual dumps occur, some file systems might
>    be dumped at level 0 earlier than a full dumpcycle.  Note
>    15 systems are scheduled to have level 0's next Mon.  I'll
>    bet some of them are "promoted" from level 1 to level 0
>    before then.

Understood.

>
> 4. balance is not based on the amount of data on disk, but the
>    amount written to tape (i.e. compressed).  According to the
>    amadmin output, amanda has seen 13.4GB (not 23) in the 17
>    file systems and they compressed to 7.5GB.  Its balance point
>    is shown as 1.5GB (with 5 runs per dumpcycle)

I was just giving the max for additional information.

> 5. Your largest file system (7.9GB) does not compress very well.
>    Only shrinks to 5.5GB.  That is > 70% of all level 0's combined.
>    No way to achieve balance with that distribution of file systems
>    and compressibility.  Probably, with time, the other 16 FS's
>    will balance among the other 4 dumps per dumpcycle.
>
> Can't say anything about your sanity :)

I have a team of experts working on this as we speak. I'm very hopeful :))

>
> I had a similar situation, on my SCSI based Solaris system I wanted
> a lot of disk storage for essentially unchanging things.  CD images,
> downloaded software packages, images of now-dead systems rather than
> always going back to tape, that sort of thing.  So I added a huge
> (for me) IDE disk.  It swamped my balance just like your setup.
>
> I decided to switch from backing up the entire disk to backing up
> individual top-level directories.  So now that one disk is backed
> up, not as "/w", but as /w/old-sys1, /w/old-sys2, /w/DownLoads,
> /w/cd-images, ...  Six entries in total.  One of the entries is
> still "/w", but on that one I use a special exclude-list that
> eliminates old-sys1, DownLoads, ...
>

That's very much like the large filesystem I have. And that's very much what
I'm gonna try to do -- separate the static from the dynamic data. It's
mostly Samba shares, so I'm working out the details (logon scripts, groups,
drive mappings, etc)

Thanks for the spot on response,
Doug

P.S. I apologize if this email came thru to the list as HTML. M$ Outlook was
"thinking" for me again ;>

Reply via email to