Memory Utilization increasing on Publisher

[TheCardMan]

Using real time monitor we noticed issues with the PUB and SUB consistantly at 98% Memory utilization. Looked into it and found out that when we upgraded to CCM 4.0 months ago we should have also upgraded the amount of RAM in the boxes. They are 7835 servers running 1gig Ram each. So we upgraded the SUB to 3 gig and the Pub to 2 gig. The utilization reading were now down to 25% for the SUB and just under 50% for the PUB. Noticed that the PUB was slowly increasing over a few days and is now at 70%. If all the devices are registered to the SUB normally, what could be causing the PUB to rise in memory utilization?

 

[jerake75]

That memory increase is normal for CCM 4.0 and above. I believe Cisco has attributed it to SQL server usage, how it grabs memory slowly over time, but will release it as other processes require it (Cisco's words, not mine).

Incidentally - just a thought - why did you give your sub more memory than your pub? Your pub runs more processes...I always double the memory in the publisher (4GB) and put 2 GB in my subs, no high memory utilization.

Also, what exact version of CCM are you running (4.0.1? 4.0.2a)? Not too familiar with 4.0, since I know it was extremely buggy and I skipped that version for my production boxes, but I think there was a memory leak associated with one of the 4.0 versions...

TMH

 

[TheCardMan]

We are using 4.0(2a)sr 1a. Noticed that the big memory user in the Pub was SQL. We figured that the SUB does all the device registering and we gave that more. I am adding an additional gig to the PUB tomorrow so they will both be at 3 gig. Which version did you see they had memory leak issues with?

 

[MG2005]

I know the RTMT was errantly reporting high memory and CPU usage in many versions of CCM (known Cisco bugs), inclusing 4.1.2 SR1 which i had been previosuly running. i recently upgraded to 4.1.3 SR1 and applied a patch that has since remedied this issue..

Memory leak is another possibility but I dont believe it was as common as teh errant reporting

the SQL explanation was another possibility

2GB of RAM is a new Cisco minimum for all servers in a cluster I believe

 

[jerake75]

Just glancing through the Cisco bug toolkit, it looks like all the memory leaks were in 4.0.1, not 4.0.2. And wow, there were a lot of them.

So it must be that sqlserver.exe process taking up your memory on your publisher...

TMH

 

[MG2005]

CSCeg41834 Bug Details


Headline RTMT LowAvailableMemory Alert shown when CM is operating normally
Product ciscocm
Feature OTHERS Duplicate of
Severity 3 Severity help Status Resolved Status help
First Found-in Version 4.0(1) First Fixed-in Version 4.0(02a)SR02, 4.1(2)ES29, 4.0(02a)ES37 Version help
Release Notes

Symptoms:

RTMT shows a LowAvailableMemory alert, indicating that there is less than 10%
free physical memory on the system (using the default threshold). This usually
occurs for the publisher.

Conditions:

Using RTMT with a CallManager 4.0 cluster

Workaround:

First, verify that the system is not showing symptoms of a memory leak.
-Check that the Committed memory usage does not greatly exceed the amount
of physical memory in the system.
-In most cases, the Total Commit Charge as shown by Task Manager will be
only slightly higher than the Physical Memory Total shown.

Once having verified that there is no memory leak issue, any of the following
workarounds could be used:
-lower the threshold for the LowAvailableMemory alert to a lower percentage
OR
-disable the alert
OR
-disable the alert and create a new alert based on "% Committed Bytes in Use",
perhaps at a threshold of 70%, depending on the system in question.

Note: you also need fix for CSCsa76784 to completely address this