Metrics / Span Correlations #123
Conversation
text/xxxx-metrics-correlation.md
Outdated
|
|
||
| # Open Questions | ||
|
|
||
| * Metric vs span tags |
There was a problem hiding this comment.
should hopefully be addressed by https://github.com/getsentry/rfcs/blob/main/text/0116-sentry-semantic-conventions.md - let's just go all in on OpenTelemetry conventions
text/xxxx-metrics-correlation.md
Outdated
| processor = Processor() | ||
|
|
||
| # This creates a span with op `timing_measurement` | ||
| with metrics.timing("processor.process_batch"): |
There was a problem hiding this comment.
Let's use the naming scheme of span name/metric name instead of using op (aligns with otel, removes educating about op types, aligns with other metrics products).
Op already has a meaning to sentry today (db.X, resource.Y), and that concept generally has caused us issues, let's not introduce it for any metrics API (we can always make op an optional thing people can set and map it to the same semantics as span op).
There was a problem hiding this comment.
I mostly wrote this up in the light of what we have today. I agree that the op/description situation is not great so ideally this can recommend something else instead.
text/xxxx-metrics-correlation.md
Outdated
| with metrics.timing("foo"): | ||
| pass | ||
|
|
||
| with start_span(op="timing_measurement", metric="foo"): |
There was a problem hiding this comment.
how does this work when we are using an otel API under the hood? We'll have to add logic in otel span processors I guess?
There was a problem hiding this comment.
Yes, that is an open question I'm not fully sure about yet. OTel Metrics themselves do not suggest any span based metrics changes. They are doing the correlations OOB with exemplars: https://opentelemetry.io/docs/specs/otel/metrics/data-model/#exemplars
But I do think the case of span + timed metric is very common and ideally we can find a way to funnel that out with otel span APIs somehow.
This RFC proposes a way in which metrics and spans can be correlated. This is a work in progress.
Rendered RFC