I run 3 nodes etcd cluster with additional flag --experimental-corrupt-check-time 1m. Flag itself does work, so I could see log messages from the Leader:

{"level":"info","ts":"2023-07-17T13:23:13.603Z","caller":"etcdserver/corrupt.go:259","msg":"finished peer corruption check","number-of-peers-checked":2}

I wished to test that feature and made following operations on the third node only:

echo "Hello World" > etcd/member/wal/0000000000000000-0000000000000000.wal
echo "Hello World" >> etcd/member/snap/db

As you can see both db and *.wal are, from now on, not consistent across a cluster.

I expect that etcd will raise CORRUPT_ALARM accoring to documentaion, but nothing actually happend I still see NO ALARMS on each node as well as successfull "corruption check" by etcd, which logs out:

{"level":"info","ts":"2023-07-17T13:23:13.603Z","caller":"etcdserver/corrupt.go:259","msg":"finished peer corruption check","number-of-peers-checked":2}

Please also note that I run 3 nodes cluster and all 3 are healthy and do respond. But "number-of-peers-checked":2 is still 2, though...

0

There are 0 best solutions below