galactic-router reflector logs a service route map error every second #834

Summary

galactic-router in route-reflector mode (GALACTIC_ROUTER_REFLECTOR=true) starts the private-service route controller. Its startup sync retries every second against the node-local service route map, and that map only exists on nodes that run galactic-cni. Route reflector nodes don't run galactic-cni, so the sync never succeeds. The router logs one ERROR line per second for as long as the pod runs:

ERROR	initial service route policy map adoption and sweep	{"error": "initialize service route maps for startup sync: serviceroutemap: open service_route_table: no such file or directory"}

That is about 86,000 lines a day per reflector. Routing is unaffected: BGP sessions, EVPN reflection and the other controllers behave normally. The cost is noise in the log and alert pipeline on every reflector.

Where

• cmd/galactic-router/root.go#L265 sets up the service-route controller in every mode, reflector included.
• internal/controller/serviceroute_controller.go#L296-L321: the startup sync logs at ERROR and retries on a fixed one-second timer. It has no backoff and no limit.

This started after v0.5.4, with the private service route work in #634.

Reproduce

1. Deploy config/galactic-router/overlays/control/ on a node that does not run galactic-cni.
2. Read the galactic-router-rr pod's logs. The error above repeats once a second, and it resumes after every restart.

Expected

The reflector programs no node-local datapath, so it should not start the service-route controller (or at least its startup map sync). Separately, on plain-mode nodes the retry should back off, and a missing map should log below ERROR until it has been missing for a while. The comment at the setup site says the datapath loader may legitimately start after the router.