Has anyone got experience of using Application Profiling in reasonably large scale live environments with v11.0.1?
We are trying to track down a potential rogue process which seems to be causing a database server to freeze up from time-to-time. The symptoms are that all users (approx 400) see their connections freeze up (in both our applications and Sybase Central etc - also new connections are refused) for 10-15 seconds before things return to normal (until the next time). While the freeze up is happening one CPU core goes to 100% and the other 17 go to zero (normally activity is well spread between cores). When the freeze up finishes most users carry on where they have left off - but some are disconnected.
We have not been able to associate the times of the freeze-up with any known user activity or timed event. So, sensibly enough, SAP Support have suggested using Application Profiling to try to identify the process that is causing the issue - as a first step to resolving it.
However our experience in v10.0.1 was that while Application Profiling was fine to use in test (or very small scale production) environments - in busy live environments the servers were very quickly overwhelmed by the extra load placed on it (even when an external tracing database was used).
Unfortunately we cannot limit the Scope of the profiling since we have no idea where the problem is coming from. Also I'm not sure what would be appropriate settings for type and conditions when we are (at this stage) just trying to identify the problem process with minimum impact. I'm worried that just finding long running process won't help, since during the "freeze-up" all process will appear to be long running!
v188.8.131.5213 Win Server 2008 R2
asked 30 May '13, 11:32
I have used it with an external tracing db for short times like 30 minutes or so, without seeing any noticeable degrading effect for the clients.
I have had a similar problem with 11.0.1 see: a sequential scan blocks all new connections
answered 31 May '13, 03:30