VCS 5.1 and Informix 7.31

[forest1991]

We have upgraded to VCS 5.1 recently and use the 5.1 Informix agent to monitor our DB. Since the upgrade the DB query times have slowly crept up so we know experience performance problems.

We have tested stopping the agent and the query times increase. Stopping VCS seems to return the DB query times to their expected values.

Has anyone seen this type of issue before?

Thanks

[mikebounds]

This is very strange as VCS and the agent should have no effect on Informix.  The Informix agent monitor only looks at processes, or does an onstat if you have SecondLevelMonitor (it can run a custom MonitorProgram, but I am guessing you are not using this attribute), so it should not effect query times.

Also you say when you stopped agent (do you mean you ran haagent -stop) the times increased - did you mean to say decreased (but not to original levels).

Stopping VCS should also have no effect, because VCS is meant to have a light footprint - you should check the had daemon and all agents (processes shown by ps -ef | egrep "VRTSvcs|had") to check they are not using excessive CPU or memory. 

I have recently upgraded a VCS cluster from 5.0 to 5.1 with Informix agent and DBA's have not reported any slowdown.

You should check your VRTSacclib is at the version (or higher) than specified by the Informix agent version you have installed, but even if this were wrong, I would be surprised if this would effect performance of DB.

Mike

[forest1991]

In answer to your questions, when we stopped the agent our db quesry times did increase, yet we have very many other external factors so this could have been a coincidence.

All the VRTS processes do not seem to have a excessive footprint.

We upgraded from VCS 4.1SP2 to 5.1

We have another cluster running VCS 5.1 and Informix 9.4 since last may and have no issues. Current versions of agents -

VRTSacclib    PSTAMP:  Veritas-5.2.2.0-11/20/09-17:57:27

VRTSinformix  PSTAMP:  5.1.0.0 200909231558

[mikebounds]

How many times have you stopped VCS (presumably "hastop -force) and saw dbquery times decrease - could this be coincidence as well?

Mike

[forest1991]

We used hastop -local -force

We stopped this morning at 10 and then started at 13:00, stopped again at 15:00. The second time we stop we seem to be experiencing different behaviour.

But we will have to collect some more stats for the next few hours.