Raised by the review of PR #305 (issue #272) and left undone deliberately.
What happens
The workspace projection sweeper runs every 30 seconds. When a configured server is unreachable it logs a WARN and moves on, which is the right behaviour for a background reconciler. It then does the same thing 2879 more times that day.
What it costs
A log that repeats an unchanging fact every 30 seconds is a log nobody reads. It also buries the events that matter between it.
What would close this
- Log the first failure per server, then stay quiet until the state changes, and log the recovery.
- Or back off: 30s, then 1m, 5m, capped, resetting on success.
Either way the reader should be able to tell how long a server has been unreachable from one line rather than by counting.
Source
Integration review of PR #305.
Raised by the review of PR #305 (issue #272) and left undone deliberately.
What happens
The workspace projection sweeper runs every 30 seconds. When a configured server is unreachable it logs a WARN and moves on, which is the right behaviour for a background reconciler. It then does the same thing 2879 more times that day.
What it costs
A log that repeats an unchanging fact every 30 seconds is a log nobody reads. It also buries the events that matter between it.
What would close this
Either way the reader should be able to tell how long a server has been unreachable from one line rather than by counting.
Source
Integration review of PR #305.