Multicluster installation - cross-cluster routing connection failure

Hi, I am setting up a Shared control plane (multi-network) installation and followed the official documentation multicluster/shared-gateways. The mesh include 2 clusters: GKE(primary - cluster1) and EKS(secondary - cluster2). I am usng istio-1.4.3

All pods, services and ingress gateways are running as expected but when I test the cross-cluster routing with: kubectl exec --context=$CTX_CLUSTER1 -it -n sample -c sleep $(kubectl get pod --context=$CTX_CLUSTER1 -n sample -l app=sleep -o jsonpath='{.items[0].metadata.name}') -- curl helloworld.sample:5000/hello I get the error:

upstream connect error or disconnect/reset before headers. reset reason: connection failure

Is there a policy that I need to apply or any other changes that is not mentioned in the documentation?

I have the same issue. The service in the remote cluster is working properly when hit from the local svc.cluster.local endopoint. But the cross-cluster traffic gives me the same error.

*   Trying 240.0.0.2...
* TCP_NODELAY set
* Expire in 200 ms for 4 (transfer 0x563d8e483680)
* Connected to httpbin.bar.global (240.0.0.2) port 80 (#0)
> GET /headers HTTP/1.1
> Host: httpbin.bar.global
> User-Agent: curl/7.64.0
> Accept: */*
>
< HTTP/1.1 503 Service Unavailable
< content-length: 91
< content-type: text/plain
< date: Fri, 28 Feb 2020 16:40:22 GMT
< server: envoy
<
* Connection #0 to host httpbin.bar.global left intact
upstream connect error or disconnect/reset before headers. reset reason: conne/

I’ve tested connectivity through out both cluster, and it’s working properly. The problem appears when I try to access the remote service using a Service Entry.

I see the traffic arriving in the remote ingress gw, but I don’t see anything in the envoy sidecar of the application pod.

Btw, I’m using port 15433 in my local cluster service entry. As long as I understand, traffic coming from an envoy sidecar to that ingress gw in a remote cluster, can be routed in the remote cluster without the need of adding VS, DR or GW in the remote cluster. (at least that’s what the docs say.)

Any ideas?

Is there any update on this issue? I am facing the same problem.

I am also facing the same issue. Please share the steps if you have found the solution…

I am also facing this problem. Anyone solved it yet?