Update resource detection mapping to be compliant with metric exporter spec
psx95 opened this issue · 4 comments
For MR types cloud_run_revision
and cloud_function
, the metric exporter spec currently falls back to either generic_task
or generic_node
.
This is currently being violated here. These explicit mappings need to be removed in order to be compliant with the spec.
The resource detector is in a weird state because the upstream one is much older and needs to be updated to wrap this. This should probably happen before that.
After taking another look at the code, I believe we have the same behavior as Java already for the metrics exporter. It uses generic_task/generic_node for Cloud Run and Cloud Functions because includeUnsupportedResources
is false
here
Does that sound good @psx95?
However, for the Trace exporter we still send the old monitored resource as span labels (g.co/r/cloud_run_revision/...
labels). I think we decided to no longer do that. I will send a PR to stop doing that.
I see, I did not notice the flag variable.
It is fine in that case.
Thanks, I will send a PR in any event to stop sending cloud_run_revision
and cloud_function
unsupported MRs for the trace exporter, so this is in line with what Java and Go do.