Hey, I’m brand new to concourse and I’m trying to get it reliably running on my k8s cluster with the latest stable Helm chart. I keep running into something like the following when running a simple pipeline with the git resource (though once in a while this doesn’t happen oddly):
Get http://127.0.0.1:8080/containers: dial tcp 10.2.1.147:35417: connect: network is unreachable
I’m using MetalLB and setting the service type to loadBalancer for web.
I have the web node / atc on port 80 which is automatically forwarded to a different container port (not 35417) on 10.2.1.147.
i.e. from the concourse-web k8s service config:
ports: - name: atc nodePort: 30456 port: 80 protocol: TCP targetPort: atc - name: tsa nodePort: 32417 port: 2222 protocol: TCP targetPort: tsa
Should there be an extra port in the concourse-web Service spec for the /containers endpoint or should it be the same port as the ui/atc? I’m not sure where the http://127.0.0.1:8080 and 35417 are coming from though the former sounds like a default IP and port. This is happening even when using a clean Postgres DB and clearing out the worker PersistentVolumes so it shouldn’t be caused by stale data.