Metrics for OPA evaluation timeouts?

[ghost]

Does the OPA sidecar emit any metrics corresponding to evaluation timeouts?
When trying to debug scenarios like the one discussed here, found that metrics like builtin http latency and OPA evaluation latency are not emitted when the query execution ends abruptly. There is just an error with message context deadline reached during query execution, but not much info to confirm an evaluation timeout. A dedicated metric can help. Thanks!

PS: Alternatively, if it can be GUARANTEED that all scenarios with error message context deadline reached during query execution are due to evaluation timeouts only, the OPA decision logs might serve the purpose.

[anderseknert]

Hey there! Is that deadline not in the calling service, i.e. envoy in that case? Or is that returned by OPA?

[ghost]

The deadline is set at Envoy, but this error message context deadline reached during query execution is part of the decision logs emitted by OPA.

[anderseknert]

Alright. Not sure about guarantees, but FWIW I've never seen that error message in any other context :)

[ghost]

In that case, should OPA emit dedicated metrics for eval timeouts? Shouldn't be an intrusive change AFAICT, what's your opinion?