CPU Utilization Sticking at 100%

ppueringer

After upgrading to version 8.0.1 we have now started to have various Windows servers indicating 100% CPU utilization and a negative memory value then staying there across multiple polling engines. To resolve the issue we have to start and stop network monitoring services or remove/restore the CPU/Memory resource for the affected node.

Has only one else run into this?

I sent an email to tech support asking some questions about how SolarWinds polls and enters the information into the SQL database. The following it the response I got:

We use the OID:

1.3.6.1.2.1.25.3.3.1.2

Which should return the OID Name:

hrProcessorLoad

Which translates to the MIB Name:

HOST-RESOURCES-MIB

We calculate by the actual integer that we GET from the device. This means that we basically just pull the raw value out and display it. You can use the Engineer’s Toolset (if you don’t have one you can download an EVAL from our site), and use the SNMP MIB BROWSER to return the actual MIB value that is being reported to ORION. This will provide you with the exact data that we are pulling that you can compare to looking at task manager. This would be better than looking in the database but if you want to look there the data is stored in the Nodes table as ‘CPULOAD’. Please let me know if I can help you further.

When I use the provide OID and MIB browser tool to query the affected servers they report back with multiple CPU processor load values. I am wondering if the NPM is adding or using some other meathod to factor a CPU utilization value based on multiple returned entries from the polling?

I can look at the values in the CPULOAD table for an affected node and see the time that a 100% CPU Utilization spike happens then stays there. This is normally happening on our terminal servers during the beginning of the day logins.


Patrick Pueringer
Network Analyst
City of Bloomington, MN