db file is flocked by another process, or taking too long

[13893602744]

A 3-node Kubernetes-deployed etcd 3.5.4 cluster was using NFS storage as the backend. Suddenly, it failed to start and reported the log "db file is flocked by another process, or taking too long", after which no more logs were generated. What caused this and how can it be fixed?

[ahrtr]

It means that anther process already opens the db file. Most likely there is another etcd instance running on the same node.

[ahrtr]

To be clearer, the db file is flocked by another process, or taking too long is just a warning when it takes etcd more than 10 seconds to open the db file. But it won't affect the functionality.

it failed to start

Could you provide the complete etcd log?

[13893602744]

There are no more logs. It's stuck here and won't move.

[ahrtr]

If it gets stuck and never move forward, then it should be the reason as I mentioned in #20935 (comment)

[ahrtr]

Most likely there is another etcd instance running on the same node.