cybertec-postgresql / vip-manager / 33750494223

91%

push

github

[-] throttle repeated DCS error logs, closes #417 (#418)

While the DCS is unreachable, the checkers run into the very same failure
once per scan interval, and every error carried a stack trace on top, so a
long outage could fill up the disk.

* add a small log throttler: a repeating message is reported once, then
  once a minute with the number of suppressed occurrences, and the recovery
  is always reported; a message that changes is reported immediately
* use it for the request failures of all three checkers, for the repeated
  non-success status codes of the Patroni checker (a standby answers 503 to
  every single request) and for the lost WATCH of the etcd checker
* disable stack traces unless --verbose is given, they add four lines to
  every error without telling anything new about a refused connection
* limit the retry chatter of the etcd client to errors unless --verbose

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>

91 of 94 new or added lines in 5 files covered. (96.81%)

933 of 1028 relevant lines covered (90.76%)

1.41 hits per line