On Mon, Jun 01, 2020 at 15:09:54 +0200, zoi...@medialab.sissa.it wrote:
> documentation and in messages in this mailing list runspercycle set with
> a value of 5 and immediately next to it (like a golden rule) the mention
> that amdump will run 5 times per cycle excluding weekends?
>
> My question is: why excluding weekends? Is not possible to run amdump
> 5 times per week excluding Wednesday and Friday? Is there any
> correlation between the value 5 and the weekends?
These types of examples probably date back to the era when Amanda was
generally used to send backups to a no-changer tape drive... so you
needed someone physically present to change the tape before each new run
(and thus couldn't do new backups if the office was empty on the
weekends). Thus, these examples target getting a full dump once a week,
but assume that only 5 tapes will be used over the course of that week.
But indeed there's no particular reason to only do backups on weekdays
(especially when using tape changers or vtapes, etc. which don't need
daily manual intervention).
In fact, at my site the current setup (which uses vtapes) runs amdump
every night, and has dumpcycle set to 3 and runspercycle is left unset
(which means "same as dumpcycle") -- so we're both running the dump all
seven days a week and also getting more than two full-dump cycles within
each week.
>
> Another related question: In a configuration with
>
> dumpcycle 7 days
> runspercycle 5
>
> during a dumpcycle (a week) do I have to run amdump in the same days
> as in the
> previous dumpcycle or in the first dumpcycle I can run amdump during
> the days
>
> 1 2 4 5 6
>
> and in the second dumpcycle amdump can be run during the days
>
> 2 3 4 6 7?
Amanda simply uses the runspercycle number to (try to) figure out how to
space out full v.s. incremental dumps over the course of a cycle -- but
this is done each run at a time, rather than there existing some sort of
overarching schedule which later runs then have to follow.
Put another way, each time Amanda runs, it gathers up the estimate
statistics for all the DLEs in the disklist, then tries to calculate an
answer to the question "Which DLEs should I pick to full-dump today, so
that no DLE is overdue for a full dump and so that the total size of the
full dumps today is close to the daily average size of full dumps?".
(You can get some insight into the calculations involved using the
"amadmin <CONFIG> balance" command.)
Since each Ananda run makes this calculation separately, if you change
the run schedule it will simply re-calculate using the current numbers
and do the best it can.
There are lots of nuances to how that plays out, but the short version
is that assuming you do keep to the same number of actual runs over the
course of the dumpcycle, you should be fine -- with the caveat that at
least when your filesystems have a fairly constant rate of changed/added
files, if your runs aren't spaced evenly then the runs that happen after
a longer delay will have more data than the runs after a shorter delay.
That is often not really a big deal, but depending on your situation it
can be undesirable (which is why many of the examples out there are
structured to try to keep the daily backup size as constant as
possible).
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
