Open mikecote opened 4 years ago
Pinging @elastic/kibana-stack-services (Team:Stack Services)
Hi Team, Thanks for considering this feature, it is very important. I am in the field already heavily using the new framework. Alert creation for user is like a mini code / feature / capability development and deployment cycle. So the ease to configure -> test -> observe -> correct cycle is very important especially when configuring the action. Example at a large customer that is integrating with and event management system via the webhook, which parses the alert to then pass on to their notification pipeline. Simulating the alert i.e. not having to deploy the alert, then cause the alert to fire (sometimes not so easy) will speed the up the alerting deployment cycle time and effort, reduce errors / missed alerts and increase the speed to value.
So as an alert creation owner, I want to simulate the alert firing to reduce the time it take for me to deploy a correct and useful alert with minimal time and debug cycle and errors.
I created a feature enhancement issue #84417 regarding an "explain" capability for the alerts underlying queries - brief discussion on this noted that this is kind of a baby-step towards alert simulation, so thought I'd reference it here.
The ability to run simulation based off of synthetic test data is important in order to test behaviour under conditions and edge cases that are relatively rare and may not be present in the actual historical data.
@mikecote and @gmmorris Can you help us to understand the features?
@mikecote and @gmmorris Can you help us to understand the features?
TBH this is more of a placeholder, as we don't quite know the feature requirements. We'd likely want parity with what Watcher provides, but we shouldn't approach this as a copy of Watcher, but rather as an opportunity to find out what our customers might actually want here.
It would be worth chatting with Product.
Ability to preview how an alert / alert type would behave with a given set of parameters. This could be based on historical or mocked data.