We have a global SLA defined for which 85% of all transactions must be less than 3 seconds. I'l like to create an Incident Rule for this but not sure how. Currently I have BT's defined for user actions and PurePaths that take longer than 3 seconds thinking I could compare it to PurePaths and User Actions in general. Any thoughts?
Alerting based on percentiles has been an open request for a long long time.
Luckily for you 90th is pretty close to 85th, especially when you consider how normal these distributions tend to look. You can hence use an "All PurePaths" and an "All User Actions" BT (with no splitting) to track and alert your 3 second rule. Here's what you do:
Now, when the 90th percentile breaches 3 seconds, you will be alerted.
Hope that helps,