On Wed, Oct 31, 2018 at 08:29:08 -0400, Chris Nighswonger wrote:
> FWIW, here is the output of amadmin balance before last nights run and
> again this morning. No overdues, so I guess that's good. I'm not
> experienced enough to make much of the balance percentages, but am now
> wondering if I should work at breaking up the large DLEs into smaller
> subsets as several have suggested.
> 
> root@scriptor:/home/manager# su backup -c "/usr/sbin/amadmin campus balance"
> 
>  due-date  #fs    orig kB     out kB   balance
> ----------------------------------------------
> 10/30 Tue   25  359009166  284972243   +102.1%
> 
> 11/03 Sat   15  632526057  420083122   +197.9%
> ----------------------------------------------
> TOTAL       40  991535223  705055365 141011073
>   (estimated 5 runs per dumpcycle)
>  (13 filesystems overdue. The most being overdue 1 day.)

Regarding "balance percentages": if you divide the "TOTAL out kB" figure
by the number of runs per dmpcycle, you'll get the number in the bottom
right corner of the chart (i.e. "141011073").  Amanda tries to spread
out the full dumps so that volume of full dumps happens each day (or, in
other words, so that the full dumps are split evenly over each day in
the dumpcycle).  The "balance" figures is simply a calculation of how
the currently-scheduled cycle compares to that ideal -- so in this case
the 10/30 figure is about twice the average (102% above the target),
while the 11/03 figure is about three times the target.

> root@scriptor:/home/manager# su backup -c "/usr/sbin/amadmin campus balance"
> 
>  due-date  #fs    orig kB     out kB   balance
> ----------------------------------------------
> 10/31 Wed    1          0          0      ---
> 
> 11/03 Sat   39 1079730080  776153947   +400.0%
> 11/04 Sun    0          0          0      ---
> ----------------------------------------------
> TOTAL       40 1079730080  776153947 155230789
>   (estimated 5 runs per dumpcycle)

Am I correct that you actually ran two separate amdump runs within the
calendar day of 10/30 (with the first "balance" command executed between
the runs)?  That would explain why all 39 DLEs are now showing as due on
the same day.

Anyway, the "clump" showing here is a direct result of the fact that all
your DLEs got set to full dumps yesterday (for whatever reason that
happened).  

Over the rest of this cycle, you should see the planner promoting about
one fifth of your full-dump volume to that day, so that the full dumps
spread back out to near a "zero" balance figure.  (That is, some level
0s will happen sooner than a full dumpcycle after the last one, to
spread things back out.)

If you know for sure that certain DLEs are larger than (or very close
to) the balance size, it will definitely help to split them.  Otherwise,
I'd say you might as well wait 5 days and then take a look at the
balance listing, to see if Amanda ended up having trouble evening things
back out over the course of the cycle.

                                                Nathan

----------------------------------------------------------------------------
Nathan Stratton Treadway  -  natha...@ontko.com  -  Mid-Atlantic region
Ray Ontko & Co.  -  Software consulting services  -   http://www.ontko.com/
 GPG Key: http://www.ontko.com/~nathanst/gpg_key.txt   ID: 1023D/ECFB6239
 Key fingerprint = 6AD8 485E 20B9 5C71 231C  0C32 15F3 ADCD ECFB 6239

Reply via email to