Upstream is missing exponential backoff which we currently have available in the agent. I think it's okay to miss out that feature for now, but we need to prioritize it upstream

Upstream is missing exponential backoff which we currently have available in the agent. I think it's okay to miss out that feature for now, but we need to prioritize it upstream

Whoops, I forgot to close that issue. Exponential backoff was added in the last PR.

so it should be applied here? because I tested and there was no backoff, it just iterated trying to connect every 30 seconds. Does it need enabling?

so it should be applied here? because I tested and there was no backoff, it just iterated trying to connect every 30 seconds. Does it need enabling?

Hmm, that's odd. It's set here in the opamp manager. Though maybe there's an issue with the implementation itself. Just for the record, the upstream retry mode is only enabled when the server returns an unavailable error status or when the http response code is either 503 or 429 (assuming that the server doesn't return a "suggested delay" value). Those are the only scenarios where the exponential backoff should kick in. I'll double-check in case there's an issue around it.

ah, I test with an unavailable port, so connection refused

ah, I test with an unavailable port, so connection refused

Got it. Still, I took the time and added a test case for exponential backoff in OpampManagerTest just to make sure it works well even after potential upstream changes in the future, as it's still incubating.

I added open-telemetry/opentelemetry-java-contrib#2274 which we'll get here when it winds onto the version that has it