Performance drop when using external events about durabletask-netherite

[sebastianburckhardt]

I have looked at the traces and one of the issues I am seeing is that there are issues with how RaiseEvent is currently implemented:

• it is inefficient because it actually performs two roundtrips to the partition (first calls GetInstanceState, then calls RaiseEvent) instead of just one. This means it has high latency and wastes EH throughput.

• there is a consistency issue because the activity is using a client when calling RaiseEvent, and this client may see an older state (where the instance does not yet exist) because GetInstanceState may see a slightly stale instance state.

The best solution to both of those problems is to fix the internal implementation of RaiseEvent so it uses a single, strongly consistent round trip to the partition.

[sebastianburckhardt]

This is not necessarily expected, no. If you can share the app name and the time window you used for these two tests, I can take a look at the internal telemetry and perhaps figure out what is going on.

[lecamarade]

Thank you, I'll log a support request. I wanted to make sure there was nothing in the architecture of Netherite that would explain it.

[cgillum]

@sebastianburckhardt can this issue be closed?

[sebastianburckhardt]

From what I understand this was fixed. Closing the issue.