Add new routes by group instead of independent routes

Is your feature request related to a problem? Please describe.

Related to how SCG persists (add + refresh) routes dynamically via Actuator endpoints

Currently, the routes are added one by one using /actuator/gateway/routes/{id}. In order to get changes applied, we need also to call /actuator/gateway/refresh endpoint which iterates over all the RouteDefinitions to refresh the Route instances.

For example, adding two routes requires the next requests

• POST /actuator/gateway/routes/my-route-1 -> add my-route-1 into RouteDefinitions
• POST /actuator/gateway/routes/my-route-2 -> add my-route-2 into RouteDefinitions
• POST /actuator/gateway/routes/refresh -> convert my-route-1 and my-route-2 to Routes

When a route cannot be instantiated (for example, a predicate with wrong arguments), there are some issues depending on the spring.cloud.gateway.failOnRouteDefinitionError flag

When spring.cloud.gateway.failOnRouteDefinitionError=true

• The wrong route remains in the RouteDefinitions, preventing new routes to be applied because the /refresh will fail silently when the wrong routeDefinition is again processed

When spring.cloud.gateway.failOnRouteDefinitionError=false

• There is no feedback about the wrong route and the only way to figure out which route is wrong is by sending a request for getting all the routes and checking which wasn’t applied

Additionally, it also has the following issues

• The /actuator/gateway/routes/refresh replies 200 OK even if the route failed
• It cannot treat routes as a whole group. If one route is wrong, I don’t want to get routes available yet

It is extremely worth having feedback about which routes are wrong and refreshing them as independent groups. Then, I can update SCG routes incrementally without breaking the current RouteDefinitions I have.

Describe the solution you'd like

To use a metadata label for identifying groups of routes in SCG

• If one route fails during /refresh, the whole group shouldn’t be applied
• If no group, the route is treated as “default” group, having the same behaviour as in the past

Describe alternatives you've considered

I considered other options

A. Adding a query parameter to the /refresh endpoint so I can refresh specific route groups

• But it doesn’t work because CachingRouteLocator rebuilds all the routeDefinitions, so I cannot incrementally refresh new routes

B. To implement a new endpoint that receives N routeDefinitions and validate them

• The routeDefinitions are converted to Routes
• Any error is returned in the response with the proper error code
• If routeDefinitions are valid, the routes are stored in the RouteDefinitions so the next /refresh will apply them as usual

Additional context

Example expected behaviour having SCG running with management port 8090 I would like to get the next response after running the script below

Routes persisted:
"route-1-1"
"route-1-2"
"route-3-1"

Only group-1 and group-3 have a valid configuration.

add_route() {
# ... it depends on the solution, I can share my implementation using groupBy metadata approach
}

add_route "group-1" "route-1-1"
add_route "group-1" "route-1-2"
add_route "group-2" "route-2-1"
add_route "group-2" "route-2-2"
add_route "group-2" "route-2-3" "wrong-uri"
add_route "group-2" "route-2-4"
add_route "group-3" "route-3-1"

curl --location --request POST 'http://localhost:8090/actuator/gateway/refresh'

echo "Routes persisted: " 
curl -s "http://localhost:8090/actuator/gateway/routes" | jq '.[].route_id'