SRV lookup timeout from golang on kubernetes / coreDNS

Question

Trying to make percona operator work:

https://github.com/percona/percona-xtradb-cluster-operator/blob/main/cmd/peer-list/main.go#L50

The golang code do a simple SRV lookup to get all pod IPs behind a service:

_, srvRecords, err := net.LookupSRV("", "", svcName)

This give error:

2021/11/12 19:55:22 lookup XXX.XXX.svc.cluster.local on XXX:53: dial tcp XXXX: i/o timeout

But:

• the pod can access to the DNS server (curl is OK)
• I set DNS timeout to 10s via container dnsConfig
• resolution is working well dig srv XXXXX.XXXXX.svc.cluster.local is OK

Answer 1

The issue was network!

Our infra guys have disallowed TCP traffic to the kube DNS server, they was thinking DNS need only UDP :/ which is wrong !

Kafka clustering also need DNS over TCP.