cancel
Showing results for 
Search instead for 
Did you mean: 
Create Post

FEATURE REQUEST - Capture top x number of processes by CPU/Memory at polling vs using alert triggers/executables after the fact.

FEATURE REQUEST - Capture top x number of processes by CPU/Memory at polling vs using alert triggers/executables after the fact.

Currently if we want to poll the top processes by CPU during an alert, it must be done after the alert has already been triggered by running a separate executable as a trigger action.  This requires a bit of a delay which often sees these processes change frequently and the information provided prove to be unuseful.  For example I have CPU alerts that constantly trigger, regardless of my best efforts to capture not only % cpu time but also cpu queue length and number of time the condition exists to limit false positives, but when I attempt to capture the culprits consuming the CPU, the list of processes that come back do not total anywhere near the amount of CPU % being reported. 

As such it would be nice to have a feature added to the monitoring templates to not only capture the performance counters, but to also pull the top x number of processes AT time of polling.

3 Comments
Level 18

You can do this with a set of SNMP-based Universal Device Pollers (UnDP). They would collect the running processes every xx minutes. If you start at hrSWRun (1.3.6.1.2.1.25.4) you'll find the specific OIDs you need. Collect several, display as a table, and you'll have the data collected in the database for reporting.

Level 9

Thanks for your reply Leon.  Honestly I really don't know what that means, but it sounds convoluted.  All I know is we're using agent based polling on our Windows servers and simply wish that information could be pulled as a component monitor like all our other components.  I also don't want a separate polling just to get that information.  I partially suspect Solarwinds itself is part of what's spiking the CPU given how fast the % is reset a minute after the alert triggers, so I'm not sure if having a poll running every minute will help here.  I need this information at the exact time the CPU % itself is being polled.

Also it should be noted, I put in this request at the behest of Solarwinds Support team due to the issue I'm having.  They did not suggest anything of the sort, which is really what leads me to think this is not the solution I'm after.

mmcdonaldion​ I completely agree with you on this. I have exactly the same issue. Having agents poll our servers, and being able to view the services/processes running on each server, I'd also like a quick and easy way to be alerted with valid data. I know we can add each of the processes to be monitored, for each of the servers, but I'd think that would require a bunch of resources. Additionally, we would have to know if any new services/processes start running, and then start monitoring those as well. All of that works if you know you want to monitor specific, and existing, processes. I do not think, however, it works very well for monitoring the things you/we do not already know.