I've been working on building up diagnostics_channel and TracingChannel as a high-performance data source for tracing products to consume safely rather than relying on risky monkey patching. I would like to extend that effort to defining consistent events for serverless function providers to publish around the lifecycle of a request to form a trace boundary.
For the tracing use case a clear start and end event will be needed to scope the activity of that request. Additionally some error tracking event would be valuable. This could be a single event or it may be worth splitting out certain commit failure cases such as timeouts.
Is this something others feel makes sense as a part of this functions API initiative, or should standardization of this debugging information be considered separately? What are others thoughts on what this should look like?
I've been working on building up diagnostics_channel and TracingChannel as a high-performance data source for tracing products to consume safely rather than relying on risky monkey patching. I would like to extend that effort to defining consistent events for serverless function providers to publish around the lifecycle of a request to form a trace boundary.
For the tracing use case a clear
startandendevent will be needed to scope the activity of that request. Additionally some error tracking event would be valuable. This could be a single event or it may be worth splitting out certain commit failure cases such as timeouts.Is this something others feel makes sense as a part of this functions API initiative, or should standardization of this debugging information be considered separately? What are others thoughts on what this should look like?