Properly create telemetry factory when initializing the lambda collector#2122
Conversation
|
Maybe consider adding a test running the collector? |
I'd love to! In a previous discussion it turned up that we (otel) don't currently have any organizational ability to deploy to a test lambda container in CI (billing, auth, etc. etc.). I've asked around internally (Splunk/Cisco) about that but it's always been a low priority to tackle. So I kinda of have to use Splunk CI to test things. 😞 We could fake out the registration server (I have a codex-generated stub for that) to test locally and it would have caught this but wouldn't be as good as a proper deployment to a real lambda environment. |
|
re adding a test, looks like a pendant to issue #2079 |
|
In case it needs to be called out more explicitly, this bug is preventing the lambda collector from starting, which prevents lambda functions which use the collector from starting, which breaks applications, which is a really bad outcome for an instrumentation system. |
|
Thank you for the fix @johnbley |
Per open-telemetry/opentelemetry-collector#14003
otelcol.Factories.Telemetryis now a required field. I noticed this because our (Splunk/Cisco's) CI test was failing because the lambda collector failed to start, with the error messageThis adds the necessary telemetry factory to both places it should be initialized, and additionally drops the now-defunct
telemetry:metrics:addresssection of the configuration yml, because that was also causing a failure-to-start. In my local test the extension now properly starts.