etcd v3.4.13 compact failed after defrag

[nomorer]

What happened?

I am running etcd V3.4.13 as cluster used by a large K8s Cluster, which db size over 4GB.
Etcd compaction is triggered by the api-server every 5min by default.

Why sometimes "finished scheduled compaction" is not showing up in etcd's log , after then Etcd request latencies increases (over 1s) when compacting

What did you expect to happen?

the compaction can be finished expectedly

How can we reproduce it (as minimally and precisely as possible)?

run a defrag for a endpoint.
it is unexpected behavior

Anything else we need to know?

No response

Etcd version (please run commands below)

$ etcd --version
etcd Version: 3.4.13
Git SHA: Not provided (use ./build instead of go build)
Go Version: go1.16.6
Go OS/Arch: linux/amd64

Etcd configuration (command line flags or environment variables)

No response

Etcd debug information (please run commands blow, feel free to obfuscate the IP address or FQDN in the output)

No response

Relevant log output

No response

[nomorer]

@tangcong

[lavacat]

run a defrag for a endpoint.

Are you running with --experimental-backend-bbolt-freelist-type=hashmap?
If yes, there is a bug that causes reset of bolt db options after defrag

#13713

[nomorer]

yes.
compare to before, latency slow down significantly during compactions after the PR merged.
but etcd compaction is still abnormal, it doesn't print "finished scheduled compaction"