I was wondering if anyone had the flag to enable logging of the Spectrum SNMP Action within Introscope?
For the general SNMP Alert action, it looks like we would add the following to the IntroscopeEnterpriseManager.properties file, but I can't seem to find the one specific to the Spectrum SNMP Action:
log4j.logger.Manager.SNMPAlertAction=VERBOSE, console, logfile
We are facing an odd issue where the var bindings are getting altered, so the Alarms do not tie to the correct model in Spectrum. Any help would be appreciated. Thank you!
If there are any APM Support guru's out there, this relates to Case 00275460
Looking at your case, I see the ticket is with the Spectrum not APM resources. The APM documentation provides general directions on setting up SNMP Alerts SNMP Adapter - CA Application Performance Management - 10.1 - CA Technologies Documentation While I have seen APM-Spectrum integration documents in the past, I do not recall one that works with later releases.Usually this is done by Professrvices.
Adding hikodavis Guenter_Grossberger musma03 KulbirNijjer Who should be able to provide additional guidance.
It's tricky because it's specific to the Spectrum Integration Action. The traps will send, but it looks like the incorrect agent name is sent in the bind.
For example, to test out we set the threshold extremely low on an Alert so that it triggers. In the example below, we would expect each Agent to get an alarm; however, they populate under one model in Spectrum.
Here is the Spectrum Action.
Here is an test alert, we set the thresholds low so that it triggers.
So we would expect to get 1 alert for each server/JVM (02Node01, 01Node01, 03Node01); however, 2 alerts are tied to one model, and 1 alert to the other.
In capturing a TCPDUMP, we can see that the agent names are in fact the models in Spectrum; thus, it may be that APM is sending the wrong agent name in the trap. If anyone else is facing this isse, please let us know.
I can tell you that the SNMP action is outdated and needs to be fixed. I was working on that, but got busy. I hope either I or the Spectrum team will get something out soon for testing.
Best way for any logging surrounding this feature is to turn on DEBUG. There is currently no specialized logging for Spectrum alert actions.
However in reading your other update, you have notify by individual metric checked. You also have the trigger "whenever severity changes". So in your test above, if one metric goes above 70, the alert will be sent to Spectrum. And that is fine. Now when that same metric goes below 70, a clear will be sent to Spectrum to clear the event. However, if metric A from Agent A goes above 70 and generates a critical alert on Spectrum, it will stay there until it is cleared. So let's say 1 minute passes after metric A went above 70. Now metric B from Agent A goes above 70, thus generating another critical alert to Spectrum, then you will see 2 critical alerts in Spectrum from Agent A.
Thank you for your response. We did enable debug to see the bindings and the issue is around the fact that there should be a total of 3 total Critical Alerts: one for each JVM. However, the var bindings are showing them as going to one agent.
I'm not sure why Spectrum would not be showing 3, unless it groups them from the same agent name. In this case I would recommend opening up a Support issue.