Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-42857

Issue accessing Service NodePort/ExternalIP with externalTrafficPolicy: Local and part of Service Mesh

XMLWordPrintable

    • Important
    • None
    • False
    • Hide

      None

      Show
      None

      Description of problem:

      • the namespace the application is deployed to is member of the service mesh
      • there are following configs in the deployment of the app
        traffic.sidecar.istio.io/excludeInboundPorts: <port>
        traffic.sidecar.istio.io/excludeOutboundIPRanges: <ip_list>
        traffic.sidecar.istio.io/includeInboundPorts: '*'
        traffic.sidecar.istio.io/includeOutboundIPRanges: '*'
      • there are other network policies added to the namespace beside the two netpols that are created by the service mesh by default
      • the namespace is using EGRESS_IP assignment
      • the service type is NodePort
      • the network config is set to routingViaHost=true
      • the externalTrafficPolicy of the service is local -> in this scenario accessing the application from the F5 loadbalancers doesn't work
      • Connectivity issue between external F5 Loadbalancer is accessing http health endpoint on NodePort 30442
        F5 LB IP 1 to NodePort 30442
        F5 LB IP 2 to NodePort 30442
      • UTC timestamp (09:14)
      • pcaps (in comment including IPs)

      Net Policy with 0.0.0.0/0 from https://access.redhat.com/solutions/6999225 didn't solve the problem.

      kind: NetworkPolicy
      apiVersion: networking.k8s.io/v1
      metadata:
      name: allow-ingress-from-outside
      spec:
      podSelector: {}
      ingress:

      • from:
      • ipBlock:
        cidr: <address/<mask> --> Add the necessary external network addresses that access the respective services
        policyTypes:
      • Ingress

      Version-Release number of selected component (if applicable):
      4.14.

      How reproducible:

      Local reproduction of the problem failed so far due to the complexity of the customer's setup

      Steps to Reproduce:
      n/a

      Actual results:
      Accessing the service from the F5 Loadbalancer doesn't work (can't be checked via /health endpoint

      Expected results:
      Service should be accessible

      Additional info:

      Affected Platforms:
      Customer is running on VSphere

      Data attached

      • Service Mesh Must Gather (servicemesh-must-gather.local.448886470022687832.tar.xz)
      • Regular Must Gather (must-gather.local.6749618347009820290.tar.xz)
      • Namespace Inspect of the affected pod that can't be reached (<ns>-inspect.local.9057403696363463024.tar.xz)

              bbennett@redhat.com Ben Bennett
              rhn-support-afaulhab Anne Faulhaber
              Anurag Saxena Anurag Saxena
              Votes:
              1 Vote for this issue
              Watchers:
              4 Start watching this issue

                Created:
                Updated:
                Resolved: