Large number of cluster view changes in previously stable cluster

EventStoreDB
1

We have a 3 node cluster that has been running reliably for months now, currently on version 4.0.1.

Recently, we’ve started seeing a lot of “connection dropped” on the clients connected to it, and cluster view changes in the EventStore server nodes due with gossip failed messages.

Is there any known reason why this would start to happen? Is there a limit on the number of open network connections or similar?

Cheers,

Kristian

2

Usually this is caused by timeouts.

3

The nodes have 1 reserved CPU each (this is our test cluster) and are not maxing out on that - timeout setting left as per default. Are there any other flags to tweak, GC settings, max network connections?