Wanda,

Apologies if this was already mentioned (I haven't read this entire thread), 
but Expiration can have a dramatic impact when the log is pinned.  It does not 
in itself typically pin the log for long periods of time.  However, it can 
exacerbate the situation where the log is pinned, especially when you have 
logmode=rollforward.  The log is actually circular and has a "head" and a 
"tail".  The oldest open transaction pins the "tail" of the log.  Any other 
updates, whether they remain open or not (for rollforward) will move the "head" 
forward.  When the "head" catches up to the "tail", then the log utilization 
reaches 100%.  The faster the head moves in a situation where the tail is 
pinned, the faster your log utilization will near 100%.  Expiration is one 
thing which can move the head the fastest (IMHO).  When I see our logutil 
rising, the first thing I check is to see if expiration is running, and if so 
cancel it.  This will slow down the rate of increase in logutil.  It would be 
nice if this happened automatically, even before db writes are slowed down, but 
I don't believe it does.

This may not apply for logmode=normal, I'm not really sure.

..Paul

At 04:17 PM 5/15/2011, Prather, Wanda wrote:
>Am I correct that scheduling of background processes is important as well?
>Don't expiration, migration, backup stgpool also create log transactions?
>I'm wondering if it is the housekeeping, rather than the number of client 
>backups, that is sending my server over the edge.


--
Paul Zarnowski                            Ph: 607-255-4757
Manager, Storage Services                 Fx: 607-255-8521
719 Rhodes Hall, Ithaca, NY 14853-3801    Em: psz1 AT cornell DOT edu