Hi Niclas,
> So I guess you use monOpenDatabases and attribute AppendLogRequests, AppendLogWaits to see this ?
> Do you also look on the waits for this spid with monProcessWaits or overall waits monSysWaits.
It's my colleague who's using asemon that saw the contention on the transaction log.
I'm not quite sure what mda tables are used to come to this conclusion.
I'll have to do more investigation.
> I would think the start point should be here to have log cache for log pages, and check the size of user log cache size and then how the IO performance is in the operating system for the logs devices.
User log cache size is 8K, and IO performanc on OS level seems normal.
As I said in my reply to Chris's comment, when I notice blocking locks in sp_who, it's not about large transactions that need to be committed.
Weblogic executes a rollback on an idle connection (so normally there is nothing to rollback) in its default database that doesn't contain tables and that has trunc log on checkpoint on.
My main question is: can it harm to put delayed commit on in that database.
Thanks,
Luc.