db file flocked by another, or taking too long

[renshunhang]

I deploy etcd with k8s, and version is etcd:3.5.11.
I'm using 3 node K8s cluster, after few days, all the etcd-cluster pods went to 0/1 Running State.

default apisix-etcd-0 0/1 Running
default apisix-etcd-1 0/1 Running
default apisix-etcd-2 0/1 Running

Etcd logs:
"caller":"etcdserver/backend.go:85","msg":"db file is flocked by another process, or taking too long","path":"/bitnami/etcd/data/member/snap/db","took":"10.008086651s"}

I checked the DB file, which is about 150MB, and it doesn't seem to be in use. Why is this happening? Do I need to increase the timeout for the livenessProbe?

[jmhbnz]

Hey @renshunhang - Thanks for your question, it looks like the version of etcd you're running is a couple years old. We've merged some changes to db file flock behaviour in the last couple years, for example #17206.

Can you please confirm this issue persists with a modern version of etcd?