CDM (CPU, Disk and memory) thresholds.

Question

As a man with no prior knowledge/experience with monitoring solutions (we are using solarwinds NPM 11.5).  I have finally gotten somewhere and am now at a stage where I can start to tweak the thresholds of the environment.

I have got so far:

- All SQL servers memory and CPU have a very high threshold (97% and 99% usage) as SQL is resource hungry.

- I have run a report for all servers that gives me a huge amount of resources, plus, from the notifications that have been sent to the hunt mailbox, I have got a reasonable picture of that thresholds I would like.

My question is thus: for all of my servers, would I want to use a high usage CPU threshold (for those 'odd' times that a site is doing work/patching/other)

Or, do I set them to something more useable (eg - if A CPU had a yearly peak of 5%, I would set the critical level to 20 - 30%)

Or, is it viable to do this and leave the thresholds as is.

Any help from my confused, almost about to explode mind would be appreciated.

shaun_9999 · Answer

I have already thought about it a little more and seen a few forum posts regarding this.

The answer is to give it enough threshold so that if you get a spike for more than x amount of time, you would need to investigate.  Rather than leave the issue to slowly grow until it's broke and requires a reboot.

conguir · Answer

Hi

what we recommend is to use custom thresholds, as you are currently doing, but do not set them too low. Is it important for your environment if a server is using 30% CPU load during 40 min, even though this falls outside the normal behaviour? I guess this wouldn't be a real problem in your network, therefore you shouldn't be alerted if this happens. You should receive alerts only when something is not working fine in your environment, and probably CPU load = 30% is not a real issue.

Raul Gonzalez

Prosperon - UK SolarWinds Partners

Installation | Consultancy | Training | Licenses