Whenever a route is changed, the routing configuration is fully removed and added again on two distinct steps. Although this works well on the reload strategy, it is not the same for DCM because those two operations happen dynamically, by removing and adding the route map entry even if there is no change in the pattern used in the map.

This behavior can potentially cause some disruption on busy environments, and should lead to a short outage of the route in the case the API call to add the route back fails and a reload need to be issued.

-

Always

* Enable DCM
* Ensure log level is at least 4
* Create a route
* Change the route to a distinct weight

Logs include "del map ..." having the map entry ID of the changed route, followed by "add map ..." adding the changed route again.

Since the sequence of API calls can cause disruption, the route should reload instead whenever a route is added, changed or removed. No API calls adding or removing routes should be observed. They are logged as -v=4 and start with "running haproxy command".

Reloads can be checked from the logs or by testing the new or changed route.

Another important aspect to test is that the router is not reloaded when changing the number of replicas of the deployment. Ensure a reload didn't happen by checking haproxy pid before and after the scaling take effect and it's part of the load balance. From the router container: "echo show info | socat - /var/lib/haproxy/run/haproxy.sock |grep Pid"

-