Traefik on docker worker node isolated to cluster


I'm trying traefik on swarm and all works well, except one case, workers isolated

Traefik and services are deployed on all my nodes (managers + workers) and I'm using docker-socket-proxy on manager to check the socket docker

Unfortunately when a worker is isolated , traefik returns "bad gateway" randomly, I guess it tries to access to the service on manager and right after locally

So my question is, there is a way on workers to "remove" the unreachable distant service ?

To be more clear

If I have a manager with an http website and a worker with exactly the same website balanced by traefik on each ( and round robin DNS) I can do something if my manager is down ?


With a swarm you should have at least 3 managers for HA to avoid this situation. 5 if you want to survive more than 1 manager being offline.

Yes I known, maybe I'm not clear my question is only about an isolate node, the node still have traefik and a container (with website) running

Thank you

It's too bad, because in this case it almost working, traefik redirects to local and another workers/managers, so from users point of view, it works, don't works, works, etc, in short it works randomly

Right you either need a loadbalancer before that or some adaptive dns or a GTM to manage that.

could you fix it?

... for me, it sounds, like it "round-robins" / iterates through your distributed instances/replicas on different nodes ...

I had the same problem (with one master and two worker-nodes)