Jerry Lawson states:
>To add to what Tom says about recovery - it seems to me, based on my recent
>experience, that while it does have the trigger mechanism built in to start a
>backup, and it will terminate transactions when the log is not available
>(full),
>there seems to be some problems in this code. I am running V2.1.0.13, and have
>had a couple of occurrences of the log filling up, backups being taken,and the
>log not clearing. On one occurrence this occurred during the expire inventory
>process; on another there was a huge backup running. I canceled the backup,
>but
>the log still didn't clear after a full backup. I haven't reported this,
I'm having the same problem and reported it to IBM on March 10...I'm still
waiting for a reply. I'm running 2.1.5.13 and the problem seems to be a bit
intermittent. I say this because as you say later in your note (which I didn't
paste into this one) sometimes the database backup does, indeed, empty the
recovery log.
>because I really don't have any documentation. When I am in this situation,
>and
>I al afraid that if I take ADSM down, it won't come back up, then I not
>inclined
>to cancel with a dump.
You don't have to cancel it, just halt the ADSM server and restart it. The
recovery log will get cleared during initialization (of course, you probably
already know this).
If I may add my own experiences with this little "feature":
it seems to happen when there long running processes (run > 8 hours). This
includes things like backups, restores, space reclamation, inventory
expiration, and storage pool backups to copypools. I first encountered the
problem of the recovery log filling up at level 2.1.5.15 and backed off to
2.1.5.13. The problem went away for a while (maybe a month) and has returned.
I now restart the ADSM server on average about once per week - and I have my
logmode set to "normal"!! I did this because our ADSM server is usually
running some sort of process 24 hours per day and having it in "rollforward"
mode caused more problems (the database backup would only clear the log
once every 3 times it ran). Setting the logmode to "normal" at least keeps
the log to a reasonable size (i.e. < 100% full) longer.
I'm not sure what kind of priority the development folks have put on fixing
the bug, but it's quite annoying to have to baby-sit a server just to keep
it running.
One question: has anyone encountered this problem on V3? I've watched this
mailing list for a while and don't recall anyone ever mentioning it. Just
curious.
Jeff Renicker
IBM Global Services
|
