What is the check frequency of APM alert conditions?

Hello,

I am trying to understand how the APM alerting works in terms of timing. We currently use the condition “Error percentage has a percentage above 0% for at least once in 5 minutes” to get alerted if any transaction fails on some of our services.

This can toggle in and out of incidents for some of these as they the transactions can be infrequent themselves. We are contemplating to increase the time period on the condition to stop the repeated notifications, but we are wondering if that will effect the latency of the notifications.

I can’t find documentation on how frequently the conditions are being checked. Will that be at intervals matching the window? Or is that an independent schedule I just can’t see?

Regards,
Peter

Hi @peter.becker1

I can’t find documentation on how frequently the conditions are being checked.

Nearly all alert condition types do an evaluation check once per minute. The only exception is NRQL alert conditions, which default to once per minute, but can be adjusted by changing the Aggregation Window.

We are contemplating to increase the time period on the condition to stop the repeated notifications, but we are wondering if that will effect the latency of the notifications.

Changing your threshold time window will not affect the latency of the notifications, but it may affect the time it takes for your violations to open.

  • Thresholds using at least once in X minutes will always open a violation on the first data point that breaches the threshold, but will wait for X minutes before the violation closes, to make sure the metric is no longer breaching the threshold.
  • Thresholds using for at least X minutes will only open a violation if the metric breaches the threshold for X consecutive minutes. They will then behave as above, watching for X consecutive minutes of non-breaching data before closing the violation.

I hope this helps!

2 Likes

Thanks, @Fidelicatessen,

yes that answers my question and matches the behaviour I observed.

Regards,
Peter

2 Likes

Thank you for clarifying @peter.becker1 and let us know if you need anything else!