// From time to time, the entire registry will be deleted (see rpc.go
// RestoreEtcd). Detect this case and wipe out the registry records as
// commonly dots will be re-owned in this scenario.
if node.Node.Key == fmt.Sprintf("%s/registry", ETCD_PREFIX) {
s.resetRegistry()
return fmt.Errorf(
"intentionally reloading from etcd because " +
"we noticed the registry disappear.",
)
}
in the etcd watcher. That RPC should have a better way of telling the remaining cluster nodes to reset their registries. If this is still in use, let's change it to a NATS message, if not, don't reimplement it.
currently there's a:
in the etcd watcher. That RPC should have a better way of telling the remaining cluster nodes to reset their registries. If this is still in use, let's change it to a NATS message, if not, don't reimplement it.