Debugging why a lambda error event will not trigger cloudwatch alarm



I’m currently following the official documentation to trigger a alarm when a lambda function invocation fails.

For testing purposes my threshold is Errors > 0 for 1 datapoints within 2 minutes' which has an action ofSend message to topic notification`.

The topic exists and I can see in metrics that there is a corresponding metric delivered. I double checked the aws cloudwatch integration url, and it matches, up, but no event is created.

What should I be checking for next? I don’t see any indication that the notification reached the pagerduty endpoint

(Simon Fiddaman) #2

Hi @nikolaevarius,

What if you try a simpler / easier to generate alarm like > 10% cpu?

Do you have a valid Escalation Policy (i.e. with someone actually on-call now) attached to your Service which holds the Integration? Incidents won’t be created (they’ll be suppressed) without a valid on-call to assign them to.

I’d also check the Alerts tab (may only be applicable if your Service is configured for create Alerts and Incidents in case they are being suppressed.

Hope that helps,


Hm, I tried a > 0% CPU alarm and the alarm triggered, and I see the correposnding notification delivered without an error.

The subscription is confirmed and I have pagerduty set to always open a new incident.

No alerts are in the alert tab. And I am set to be on call for this event.

(Chiedu Uraih) #4

Hi Joe,

So we don’t share sensitive information in a public space, can I have you
email so we can dig into specific issues on this matter.

I would close off this thread meanwhile.

(system) #5