Closed sappenin closed 4 years ago
Also, see JS doc here -- CHILD peers by default neither send nor receive routes, which makes sense. E.g., a parent really shouldn't receive routes from a child. Additionally, a child doesn't need to send routes to the child because the child doesn't really care. The Child can use the parent as a default-route.
We should also consider getting rid of the AccountRelationship
since it may cause confusion. This value is only used for routing, and we could consider just using the send/receive routes instead (note: look through JS code first before removing).
See https://github.com/interledger4j/ilpv4-connector/issues/655 -- that's half of this bug.
When the server starts-up, it scans through any static routes in the database and adds these to the routing table. If the timeout on these routes is sufficiently long, this can cause the server to not actually startup and begin listening to incoming connections until all static routes have been processed.
In one instance, there were hundreds of static routes configured, each hanging on a 60s timeout. The Kubernetes cluster running this instance was set to kill the pod after 15m, so the server got stuck in a loop where it could never startup because it wasn't able to process CCP route-updates in time before being killed.
Some possible mitigations: