Open herzrasen opened 5 years ago
:+1: to documenting this, currently only in the reference.conf
: https://github.com/akka/akka-management/blob/master/cluster-bootstrap/src/main/resources/reference.conf
Trying both would improve first user experience but not sure there is a prod use case for using both for bootstrap?
I came across this because Docker Swarm's internal DNS doesn't support SRV records and we had to override the settings to get our cluster to bootstrap.
I think falling back to A records, even defaulting to A (or making it configurable) makes a lot of sense. In general, I think SRV lookups should only be necessary when non-default ports are used (which should rarely be required in Kubernetes since each pod gets an IP address).
Hi,
it would be great to (at least) clarify in the docs that by not overwriting the default values of
akka.akka.management.cluster.bootstrap.contact-point-discovery.port-name
andakka.akka.management.cluster.bootstrap.contact-point-discovery.protocol
with "" the DNS mechanism defaults to resolving SRV records, without falling back to A records. It took me several hours to figure out why ClusterBootstrap suddenly stops working. Is there a cause why no A record resolution is tried when no SRV records are found?