CUCM 8.5 Low VM Available - Recurring - What's going on?

[MitelInMyBlood]

CM running on 7835 pub with 2 subs
Approx 1500 hardphones. Most are registered to one of the subs
Approx 1200 CSF softphones (most unregistered), those that are registered are also using one of the subs.

Using default RTMT alarm settings (60% threshold) we're seeing low virtual memory alarms on the pub occurring roughly 21 or so days after a reboot. We increased the threshold first to 75% then a few days later to 80%. Now at 35 days since the last reboot we're hovering at 78~80% VM in use with periodic alarms starting to come in again and are planning to reboot tonight.

What's going on? The subs aren't having this problem, only the pub.

Thoughts appreciated!


Original MUG/NAMU Charter Member

 

[whykap]

You can try to restart the tomcat service on the publisher server only from the cli. That would alleviate the low VM, at least temporarily.
To confirm that it is tomcat using the VM type:
show process using-most memory
or something along those lines.
If it is tomcat you can restart it with this command:

utils service restart Cisco Tomcat

Only do this on the sub again.

If it is tomcat service doing this, it is a tomcat bug causing a memory leak. You will need to contact TAC to identify the exact bug.
Since you are using RTMT there are several RTMT and tomcat bugs out there, with one of most common ones being this:

http://tools.cisco.com/Support/BugT...ls.do?method=fetchBugDetails&bugId=CSCtj50884

The bug details tell you how to identify if that is the bug you are running into. Let me know if you cannot access the above link as it requires cco access, but you should have that.

 

[MitelInMyBlood]

Thanks, but Tomcat isn't the culprit. The VM pig is CCM. I got a new VAR (yippee!!) in house today and he thinks our PUB is underpowered (tell us something we didn't know) but says it's salvageable and recommends increasing DRAM on the PUB (it's currently only 4 gb)- He suggests taking it up to 32 gb and maybe even adding the 2nd processor, and setting up 16 gig for each processor, but start off throwing some DRAM at it as that may either solve it or lessen the impact, ie, give us a larger window between required reboots.


We rebooted the entire cluster last Sunday at noon. Only 4 days later (today) the VM in use is already up to 74% and climbing. The thing is like a ticking bomb.

Original MUG/NAMU Charter Member

 

[whykap]

From the devices you mentioned I don't think it is an underpowered server, more memory should help though.
Sounds like you have it under control.

 

[pndscm]

What version of 8.5? We had a similar situation that was solved by applying the latest patch. I think we were running 8.5.1.10000 before we patched it.

 

[MitelInMyBlood]

Pndscm:

System version: 8.5.1.11900-21

Planning on a meeting w/Cisco this week to discuss. Any info you can provide might prove beneficial.

Thanks!!

Original MUG/NAMU Charter Member

 

[pndscm]

The latest patch I think is 12900-7. I'd do that then see what happens. Or at least look through the ReadMe and see if there are any memory leaks being talked about.

 

[MitelInMyBlood]

Thanks. The Cisco meeting has been postponed until after the holiday. We've noticed that the memory leakage seems to increase in relation to the amount of customer data entry work, ie, the more programming we do the faster the VM in use rises. If the database remains stagnant the memory creep is much slower.

Original MUG/NAMU Charter Member