Thanks, Larry, Jagtar, Penelope, and Shahar for your input and thoughts;
hopefully I didn't miss anyone. While an errant Daylight Saving Time
process sounded the most interesting, I think the root cause in our case is
the database cleaning which NetBackup does when there are no backup jobs
running. An image DB cleaning started at 2:05 that same morning and a media
DB cleaning at 2:27. So ironically the fact that there was no activity for
a short while before an important class started, the jobs in that class was
delayed. (Usually the admins of the machines in that class are concerned
about contention, now they have a problem which seems to trace back to lack
of contention. :-)
Since the cleaning does take a while, an hour in our case, I do wonder if it
is possible to set a "Don't clean the databases during this time" flag.
Thanks again....Gerrit
Gerrit Bos wrote:
> We had a complaint from one of the server admins that a backup which
> faithfully starts at or close to 2 AM didn't start until after 3 AM.
> Resources were available, in fact, the tape drive graph show a drop to
> only one in use. Our wake-up interval is every 10 minutes, and like
> clockwork when there are active schedules, there are "find_work" entries
> in the bpsched log files. Then last night, for one hour, there are no
> "find_work" entries. A full set at 01:55, and another at 03:05, and
> none in between. Is that evidence that the scheduler stopped checking
> for an hour? If it is, can you think of any reason why the scheduler
> stopped checking for work for one hour last night? Our environment is
> NB 3.2GA on Solaris 2.7. Thanks in advance for any feedback....Gerrit
>
> _______________________________________________
> Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
|