AWS RDS swap usage is getting in alarm state.

Question

We have AWS cloudwatch alarm setup for AWS RDS swap usage.

with following configurations
if swapusage is outside the band and anamoly detection threshold is set to 2. it will trigger an alarm.

This alarm is getting triggered at unusual time from 1:00AM UTC to 4:00AM UTC from past 2 weeks.

as per this article I checked both the graphs of `swapusage` and `freeable` memory. both are constant from past 2 weeks, with little fluctuation in freeable memory. (that too in time other than above specified window)

This is very unusual time and we do not have any specific thing running at this particular time.

What can be the problem? or how should I troubleshoot this.

Answer

Hi,
you indicate that you have set an alarm with anomaly detection and that alarm triggers at an unexpected time.

An alarm changes state if the condition is violated, and that could happen unexpectedly in your case for example if:
- the data is missing, and your alarm is set to treat missing data as bad
- the data is breaching a threshold defined by the anomaly detection but you expect the anomaly detection to work differently / compute a different threshold

In order to distinguish which case you're in, can you please share the graph and history details of your alarm? When you open the history tab on the alarm details page, you can click the link in the entry corresponding to the unexpected state change to get a JSON entry that includes the values of the datapoints that the alarm evaluated.

AWS RDS swap usage is getting in alarm state.

Relevant content