Skip to content

fix(metrics): report pressure and processed counts - #285

Open
kazmosahebi wants to merge 2 commits into
mainfrom
fix/fetcher-pressure-processed
Open

kazmosahebi wants to merge 2 commits into
mainfrom
fix/fetcher-pressure-processed

Conversation

@kazmosahebi

Copy link
Copy Markdown
Contributor
  • An idle fetcher reported a scaling pressure of 40. The buffer_depth component measured the guard's used bytes against buffer.memory_limit, which defaults to 0 and was floored to 1, so any buffered byte read as full. It now measures the used share of the guard's own limit.
  • Nothing incremented records_processed_total, so the Components card showed 0 processed beside the delivered rate. Each batch now adds the rows enrichment did not refuse once the whole batch is through. An aborted flush counts nothing, so its re-fetch counts each row once, as received does.

On a dfe-docker stack the pressure reads 0.60 to 0.73, and CloudTrail reads 78 received, processed and delivered.

hyperi-ci check passes on three feature builds: the default, jemalloc alone and db-clickhouse with db-mongodb and file-tail.

The `buffer_depth` component measured the guard's used bytes against `buffer.memory_limit`, which defaults to 0 and was floored to 1. Any buffered byte read as full, so an idle fetcher reported a scaling pressure of 40.

It now measures the used share of the guard's own limit, so an idle fetcher reads under 1.
Nothing incremented `records_processed_total`, so the Components card showed 0 processed beside the delivered rate. Each batch now adds the rows enrichment did not refuse once the whole batch is through. An aborted flush counts nothing, so its re-fetch counts each row once, as received does.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant