fix: Request Flood when misconfigured cluster-cidr causes dead routes to be kept

[apricote]

Scenario

• Cluster CIDR in hccm config does not correspond to real Pod CIDRs
• Routes are created for real Pod CIDRs
• Nodes are removed again by an autoscaler
• Routes are never deleted (since fix(routes): Only delete routes in the Cluster CIDR #432) because CIDR is outside the configured Cluster CIDRs
• With ListRoutes, the existing routes are always listed again, but the gateway IPs cannot be assigned to any servers because the servers are deleted.
• In case of a cache miss we reload the complete server list.
• Since several faulty routes exist without servers, we reload the server list several times per ListRoutes

Effect

Depending on the amount of "dead routes", this can quickly exhaust the API rate limit by sending multiple GET /servers requests per second.

Solutions

• Detect misconfiguration: In CreateRoute we can validate that the Route CIDR is contained within the configured Cluster CIDR
• Slower cache refresh: Refreshing the cache multiple times per iteration is not necessary, it should be enough to refresh the cache once per ListRoutes call

[github-actions]

This issue has been marked as stale because it has not had recent activity. The bot will close the issue if no further action occurs.

[github-actions]

This issue has been marked as stale because it has not had recent activity. The bot will close the issue if no further action occurs.