chore (sync service): probablistic sampling of transaction spans #2054
+57
−1
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Part of #2032.
We noticed that most spans are descendants of the
pg_txn.replication_client.process_x_log_data
root span.Therefore, we decided to only sample a portion of those spans.
This PR introduces a custom sampler that works as follows:
pg_txn.replication_client.process_x_log_data
root spansProblem
We would like to sample all errors.
To do this we need to make the sampling decision at the end of the span when we have all attributes and events because errors are recorded using an "exception" event, this is known as tail sampling. However, the Erlang opentelemetry library only seems to support head sampling:
cf. https://docs.honeycomb.io/manage-data-volume/sample/techniques/ if you're not familiar with head vs tail sampling.
EDIT: solving this problem may require using HoneyComb's "Refinery" mechanism. So Electric would sample all traces and we would setup Refinery with custom tail sampling logic. However, this requires extra infrastructure to set up.