Socket closed during DCP replication

During cli tests for OS_CERTIFY, failures are occurring with unexpected connection refused errors.

This is happening on all *nix operating systems.

Looking at the ns_server logs, it seems to be closing itself down in the middle of the test.

Last stable build is 7.0.0-4678, and it seems to be the first defined node that crashes each time.

It's worth noting that there is a line where we call killall on memcached. I don't believe this to be the cause though, as this line has been in the codebase for years without us seeing this issue.

Install couchbase 7.0.0-4960

Create cluster with 6 nodes, one running KV, index and N1QL

Run create, update, delete and set expiry for documents

Wait for documents to expire

Get details of machine (Uses SSH to get details like OS type, distribution, architecture etc. in this case from /etc/system-release)

Test fails with connection refused

First instance of connection refused is found at 2021-04-21 03:52:08,866 in test.log