Cisco Prime 4.2 High CPU utilization

[agivens02]

I am running Cisco Prime 4.2(unix version) on a VM. CPU utilization will run between 90 and 100% constantly. Reloading the server will stabilize the CPU for a short period of time, however the CPU usage will always return to 90-100% utilization. Currently I have about 130 devices added to the system. Is anyone else experiencing this most frustrating issue?

[agivens02]

Top command from the shell shows this.

960 casuser   19   0 1661m 248m 5612 S 100.1  6.5 226:53.72 java                                                               957      

957 casuser   16   0 1267m  92m 5444 S 99.4  2.4  26:02.27 cwjava

Imagine that, Java is eating both procs. Is there a fix?

[alois.heilmaier]

Hi,

have you got an solution ?

I use PI NCS 1.3.20 (virtual appliance)

As I got a disc full alarm I have done a "ncs cleanup".

Now I have a new alarm; 100% cpu usage.

I used top to find out what is going on:
  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
26802 root      18   0 4914m 2.3g  39m S 188.6 30.5 187:57.06 java

java uses all the cpu time.

I think about a "ncs stop" and "ncs start". Is this the solution to get out of 100% cpu ?

Best redards

Alois

[agivens02]

Hello Alois,

Unfortunately I do not have a solution.  Sometimes starting and stopping processes will address this issue, other times I am forced to reboot. On the good side, once I reboot, the system seems to be stable for 2-3months before I may need to address this issue again.

At first I thought the Processor usage may be an error however when logging in through the web interface, navigating through the different modules can be quite slow. The commands below will stop and start all processes.

/etc/init.d/dmgtd start

/etc/init.d/dmgtd stop

If you acquire additional information, please share it with me.

Alan

[alois.heilmaier]

Hi Alan,

thanks for reply.
I had this problem the first time and I also had slow response when browsing the ncs gui.
The problem starts as I have done a ncs database cleanup, where the cleanup will stop and start the ncs application.
Even the cli acces was as slow as I thought I have to execute a rebbot via the hypervisor tools.

So for this moment the ncs stop and ncs start helped, but took about 30 minutes.
If I experience any further proble I will try the dmgtd start/stop.

How did you get information for dmgtd start/stop ?
Have you analysed the init script, or are there any other source to get informed ?

best regards
Alois

Sent from Cisco Technical Support iPad App

[scrye]

Hi All;

Well, I have what seems like this exact problem described by others. We are running the large physical appliance:

Cisco Application Deployment Engine OS Release: 2.0

ADE-OS Build Version: 2.0.1.038

ADE-OS System Architecture: x86_64

Cisco Prime Network Control System

------------------------------------------

Version : 1.3.0.20

Right now it is so slow as to be completely unusable. Top shows ~ 130% CPU from java!?!  WTF? (see attached)

I'll probably open up a TAC case, for now I will try the resarts recommended.

One thing I wish I could do is set the system to purge all alarms older than a certain date. Does anyone know how to do that? Right now I have > 7000 present. Been trying to delete them 500 at a time (the max allowed by the GUI), but it keeps erroring out (see attached)

All in all, our experience with NCS PI has been very, very unpleasant. It takes up a huge amount of my time babying it. Had a month-long TAC case just to get it upgraded from 1.2 to 1.3. We had data loss and had to send the backup to Cisco for massage, but even then we lost information.

Steve