Memory Leak after installing SU13

[WildCardSys]

After installing SU13 on 4.2 SCCS the server is losing RAM on a daily basis requiring a reboot once a week to recover the RAM. In comparing process with my other sites SCCS 4.2 / SU12 I have not seen any items which indicate where the memory is being used. The overall system resources appear to be normal but the RAM value continues to decline. Nortel does not have any documented cases of this phenomenon and suggests that it is something else with the server and to back out the patch to prove their theory. I am hesitant to back out a SU since it could cause other issues, since expect for the leak the server is processing calls normally.

 

[Skaret]

What type of server do you have?

There has been a problem on Compaq hardware and Symposium Web Client SW - could be the same...??

The memory leak was related to the interaction of Compaq Services installed by Smartstart (Compaq Foundation Agent and Compaq Web Agent) and Sybase.

Workaroudn was to disable these two services and reboot server.

 

[cgilmer]

You need to pull up the Performance Monitor in the Windows ultility of the Server. Set it to update only like once a minute and add ALL of the "commited bytes" from processes on the server to the chart.

You should start to see a trend of what process(s) is sucking up all of the RAM. If they give you heck about running that on the existing server you can run it on a seperate server to monitor the symposium server.

Like I said though make sure you don't have real small intervals.

Also one quick thing you could check is finding any dormant IP's on the server might be trying to outpulse scheduled reports to... that caused me heartache in the past just like you are having. It ends up spooling up the reports until finally the server runs out of memory to hold them. Rebooting only whiped out the reports until more could spool up again.

 

[rfwhite]

It's a known issue. We had the problem with SCCS on Win2K server. Your vendor or Nortel rep should be able to help. Check nitfe.exe- it may only give you trouble if you're running HDX.

 

[WildCardSys]

We are using Dell Power Edge 2450 servers for our h/w. Our support vendor says that they have spoken with Nortel and they have stated that there are no documented cases of this issue. We are not currently using any HDX script commands, however the service is active. Rfwhite could you please reference a service bulletin and/or explain how you corrected the issue on your W2K server? Thank you.

 

[cgilmer]

Did you upgrade just to upgrade or did they tell you something in SU13 was going to fix an issue you already had. If you were functioning fine on SU12 by all means roll back the SU13 to SU12 and satisfy their curiosity that it is or isn't the PEP.

If your memory leak goes away they have something they can address. If this isn't an option for you seeting up the Windows Server Monitor to log the committed bytes of the processes running on the server will nail down what is sucking up your RAM.

 

[rfwhite]

Wildcard, I'll try to get the specifics for you. In our case it was the nitfe.exe that was building from 35-40 megs as I recall to whatever memory was available. It was first noticed as a result of running HDX. If you don't see that file growing pretty rapidly it's not likely to be the same issue. It should be quite apparent if so. I don't know how much memory you have but ours was leaking away alarmingly fast- easily seen by looking at task manager snapshots a couple of hours apart.

 

[Jaimegu]

I´m using performance tool to create a history file.
%commited bytes is increasing 0.35% to 0.7% daily

 

[cgilmer]

There is a way to narrow down the committed byts to processes and I thought it was an option via the performance monitor set. You should be able to bring up each process's committed bytes and not just sever total on the server.

If what you are saying is less than 1 percent my guess is you don't have all the processes up. If its 35% to 70% I can't believe its lasting a week.

The committed bytes should just just after a reboot as the processes start up and allocate what they need to function. After about 10 minutes they should level out and only the ones that are causing a problem are those that continue to take up memory and not release it back.