I’m actually looking into how to make our sentry alerts more useful. At the moment we are only notified of new issues via Slack. But this can be hard to determine what is new and what is old noise.
From an Ops point of view we really need 2 types of alerts.
- A regular notification with a summary of events that have occurred. For us probably once a day would be enough but i think this should be adjustable.
- A alert when particular thresholds are hit. For example if we suddenly see a spike in error rates then it probably a sign of something terribly wrong. Again this would probably be another summary of the threshold window.
I don’t think we need to know every new issue as they come in unless they hey a large impact. Otherwise if they are just included in a regular summary that would work well.
BTW the Weekly summary is good for overall health but not suitable for Operations Support. It doesn’t show any useful information at a glance to determine if immediate action needs to be taken and doesn’t occur often enough.