nic-6443 commented on issue #12803: URL: https://github.com/apache/apisix/issues/12803#issuecomment-4676894757
This is the same underlying problem that's now tracked in #13141 and #13282: when upstream nodes change (including k8s endpoint updates), old targets are only marked for delayed purge in the healthcheck shared dict, and a flaw in the cleanup loop of lua-resty-healthcheck-api7 (`last_cleanup_check` is reset inside the per-checker loop in `lib/resty/healthcheck.lua`) means that with multiple checks-enabled upstreams the stale targets are effectively never purged, so they keep being probed. The fix needs to land in the healthcheck library plus an incremental target-update path in `apisix/healthcheck_manager.lua`, and that discussion is happening on the newer issues against current versions. Since this report is on 3.10.0, whose checker lifecycle code has been completely replaced by the healthcheck manager (3.14.0+), I'd suggest closing this one as a duplicate and following #13141 / #13282 for the fix. Feel free to reopen if you still see this on 3.16.0 with details that differ from those issues. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
