Two scenarios. Wade in with help, please!!!
First:
I have a WINNT implementation of NT. Large recovery log, 1 GB, Roll forware
recovery and the database backup trigger set at 90%. Robot runs out of tapes
so no more backups are possible, so the recovery log fills and finally
operations cease. Rebooting the system (the customer did this!) results in an
error when ADSM tries to restart. I have not looked at the .err log to see
why, but probably recovery log full is the reason. So I need to do a:
dsmserv extend log c:\mumblefratz\log1.dsm
to fix it theoretically. Anything else? I haven't tried this yet as the
customer has not had the time, but he had tried it before when this happened to
no avail. We wound up reinstalling ADSM and starting over. This for a number
of reasons, lest you think I've lost my mind.
Oh, the recovery log, as shown by a q log, was at 97% full when the server was
halted. Three volumes make up the recovery log, the original four MB and two
500 MB.
Second:
In the process of trying to recreate the above scenario on my lab system, I
reduced my recovery log to the original 4 MB that the installation procedure
creates. I then started a backup and sure enough I got the recovery log full
message on the client and operations stopped. The server is fine, however. I
stopped and restarted ADSM, no problem. No luck reproducing the problem on
this attempt.
So I thought to extend the log into another file to more accurately reproduce
what I had at the other site. Now I can't seem to fill the log. It will spike
up to 30% utilized, but then drop back down.
I thought the recovery log, in Roll Forward mode, would not be able to become
smaller during an operation. What am I missing?
I'm backing up the same filespaces, over and over, into the same pool on the
server to generate database and recovery log activity. I broke the servers
ability to backup the database using the trigger by marking all the volumes in
my robot private.
Any light shed on these two issues is greatly appreciated.
TIA,
Kelly Lipp
|