A few weeks ago, after my new 3.1 install had been stable for a while, I
got brave and
decided to push to Roll Forward mode and set my DB backup trigger to
50%.  I created a
device class on my optical storage device (a 3995 optical library) for
db backups and set them
to do an full every seventh time.  Everything plugged along just fine
for about a week and then failed miserably with an error that no more
recovery log space was
available.  I went back through the recovery log and discovered the DB
backup had failed but
I couldn't figure out why.

I tuned a couple of things (backup trigger, etc) but have consistantly
gotten a failure once per week when the db backup fails.  I am now
convinced that what is
happening is that the RL get to the threshold, it tries to backup to the
opticals but there are
already two platters running processed occupying the drives.  Therefore,
no drive = no backup
= full recovery log = dsmserv crash.

I thought that this RL backup was supposed to have prriority over other
processes.  Are the
processes supposed to suspend and they're not? Are they suspending but
the platter involved
aren't vacating the drives? I don't have the local disk to back up the
database to a local file system
on the fly.  Has anyone seen this?

Thanks in advance,

Hilary Whitehead
UNIX System Specialist
IBM Global Services at California Power Exchange
1000 South Fremont, Bld. A-13           Alhambra, California.  91802
Phone: (626) 537-2770                           Fax:     (626) 537-2924

E-Mail: hilarywhitehead AT hotmail DOT com