Memory Leak issue 8.4

[Pedro]

Hi All

Wondering whether anyone can help with a Cognos Planning problem we've had for quite a few months now.

We are experiencing a memory leak on our PRD environment which can't be replicated in any other environment, we've been monitoring several stats for a while now and we can see that the handle count increases; from the point when the server is restarted the handle count increases up to about 50,000, at this point we get low on memory and eventually the server falls over. The increase in handle count is more extreme the more users and / or if macros are running  however we also see the handle count increase even when there are no users on the system and no macros are scheduled.

We are currently using version 8.4.27.80

We've been told that there is a known memory leak issue with 8.4 and we should move to 10.1, however given we can't recreate on other environments and the issue never used to be a problem I think there is another issue at play.

We have tried disabling the Anti Virus software but that has had no impact on the problem.

Has anyone experienced anything like this and can offer any thoughts as to what the problem is?

Any advice would be gratefully received.
Pedro

[skinners666]

If its environmental - have you checked that the database drivers are the same versions across environments? Over the years Ive seen lots of leaky versions of DB drivers. Other than that - it may help to narrow down the issue with some more logging.
This blog may help you - http://blogs.technet.com/b/markrussinovich/archive/2009/09/29/3283844.aspx

[Pedro]

Hi there

Thanks for your response and valuable thoughts.

I can see the SQLOLEDB drivers are consistent but our DEV environment has a more up to date SQLNCLI driver, I don't think we are using SQLNCLI but it's worth a try updating the version on PRD, I know our applicaiton databases are using SQLOLEDB but perhaps our content store (or other non application databases) are set to use SQLNCLI (that would be a nice explanation).

Thanks again.

[topedgemonk]

We've been told that there is a known memory leak issue with 8.4 and we should move to 10.1

Just wanted to tell you that 10.1 has a memory leak issue of its own contributing to a performance degrade from remote locations. Other ramifications of this include users getting an error on portal when trying to initialize contributor (it fails to initialize).
While there is a fix available for it, the fix requires an upgrade of contributor as well which is very difficult if you have a global user base and is not practical to enable automatic downloads from a server situated in one country.

I've been told all this is fixed in 10.1.1 FP2  ;)

[ericlfg]

Specific problems were identified in 8.4, 10.1, and 10.1.1 in relation to the planning web service (PWS) and client system save processes taking longer than 2 minutes (120 seconds) to commit.  If the time exceeds 2 minutes then the session would be severed (and re-established seamlessly), but would cause memory to gradually increase in the epPlanningService8.exe on a machine running the PWS.

If you are indeed running into this problem, you may be able to workaround it by increasing the 120 seconds to ~900 in the epBrandingSaveResources.xml file on EACH client system.  Look for the
<Resource ID='CallTimeoutLimit"> and change the [CDATA[120]] to 900.  It is worth mentioning that even if one machine is missed, you could still run into the problem.

If this doesn't help with the problem, you may want to log a ticket with IBM support to get them to take a look.