02-13-2014 10:36 PM
Need you help urgent;
I configured Infra SPI in OM 9.0 Windows,And i deployed the SI-CPUSpikeCheck Policy on the managed node(Windows):
When I am making utilization high of the node CPU, I am getting event after 10-15 minutes but in the policy I gave pooling interval only 1 minute;
By running ovcodautil -dumpcoda | findstr BYCPU_CPU_USER_MODE_UTIL command on node but its not giving the current utilization of CPU;
I want event on the OMW within 1 minute whent the CPU utilization will be high on Managed node is there any solution possible to get the events within 1 minute;
Thanks in advance,
Solved! Go to Solution.
03-21-2014 06:18 AM
Do you have any suggestion on this issue;
I am getting CPU event after 10 minutes in Operation Manager by CPU-Spikecheck policy
Thanks in advance,
03-31-2014 07:16 PM
Binod, is your requirement to monitor every single CPU core? that's what the CPU Spike Check policy does. so in the event that your server has 16 cores there's potentially 16 alerts that you might get - so be careful while using this policy. it is designed for use with specific cases or for systems where there's only 1 or 2 CPUs and the CPU bottleneck monitor policy is considered a bit heavy weight.
Next, are you expecting an alert to be sent for a short-term spike in CPU usage? I would avoid that sort of monitoring, since it creates too many alerts in the UI and then folks start seeing too much noise in the event browser leading to unreliability.
Read this blog post about monitoring - Monitor the job queue, ascertain system bottlenecks - best practice #1
You could deploy the CPU Bottleneck Diagnosis monitor policy - also part of InfraSPI - for production uses.
Finally, yes it is true that there's a delay in the alerts - primarily due to the fact that there's 1 interval before coda data comes to monitor agent. combine this with the policy's alerting interval - and it takes two 5-minute intervals (total 10 minutes) before the alert is generated. You could use agent's real-time capabilities (cpsh, glance) if you wish to get very quick alerts, if you have the license entitlement for RTM/perfd/GlancePlus (run oalicense -get -all to check).
But again do note - you must not check for only 1 metric to ascertain a cpu bottleneck situation - you must check for multiple indicative metrics. Try opening MS outlook on your laptop especially if you large pst files - and see in MS windows task manager how CPU usage (and memory usage) climbs up during the initial startup of the software. after a while the cpu usage reduces quite considerably. keep task manager running and see how much 'spikes' you continue to see during regular course of working. it is not a good idea to send alerts to OM and the acknowledgements for each of these rise and fall situations.