nic-6443 commented on issue #12167: URL: https://github.com/apache/apisix/issues/12167#issuecomment-4676894050
This should be fixed since APISIX 3.14.0 by #12514 (commit dac0b5546). The recurring `compacted` errors happened because the shared etcd watcher restarted its watch every ~50s with the last revision it had actually seen, so on a quiet prefix the start revision could fall behind etcd's auto-compaction and etcd would cancel the watch with `compact_revision` — and then every resource type did a full readdir resync, which is exactly the CPU fluctuation you captured. Since #12514, the watcher fetches the cluster's latest revision on every watch timeout and upgrades its start revision in `apisix/core/config_etcd.lua`, so an idle prefix no longer falls behind compaction. You're on 3.10 which predates the fix; upgrading to 3.14.0+ should make both the error and the CPU spikes disappear, so I'd suggest closing this one. If you still hit `compacted` after upgrading it would mean your compaction window is shorter than the watch timeout, which would be worth a fresh report. -- 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]
