App Connect

 View Only
Expand all | Collapse all

ACE instalation on a K8s cluster - operator setup

  • 1.  ACE instalation on a K8s cluster - operator setup

    Posted Mon January 09, 2023 03:45 AM
    Hello, I am in the process of setting up a new ACE environment in a fresh k8s cluster [1 master, 2 worker nodes; CentOS; Calico CNI].

    I am going thru the steps described here - Installing IBM App Connect in a Kubernetes environment

    Currently stuck on step - Creating the IBM App Connect Operator subscription

    I've prepared my subscription resource and the expected behavior is that once I apply it, it would prepare an install plan [which I've set to require manual confirmation] after which it should install the operator.

    However this does not occur - the logs start reporting the following message:

    failed to populate resolver cache from source operatorhubio-catalog/olm:
    failed to list bundles: rpc error: code = Unavailable desc = connection
    error: desc = "transport: Error while dialing dial tcp: lookup
    operatorhubio-catalog.olm.svc on 10.96.0.10:53: read udp
    192.168.186.131:49013->10.96.0.10:53: read: no route to host"

    where 192.168.186.131 is the cluster IP of the ibm-appconnect-catalog/catalog-operator pod on one of the worker nodes, while 10.96.0.10 is the kube-dns service.

    I've received similar messages during the k8s cluster setup and the issue was the firewall block on some ports. After unblocking the required ports it would be resolved. Currently, the following ports are enabled in the cluster:

    master - 6443/tcp 10250/tcp 2379-2380/tcp 10251/tcp 10252/tcp 10255/tcp 5473/tcp 53/tcp 53/udp 179/tcp 50051/tcp
    worker - 10250/tcp 30000/tcp 10251/tcp 10255/tcp 5473/tcp 53/udp 179/tcp 53/tcp 50051/tcp

    But I am unable to detect which port is blocking the connection attempts; I've even disabled the firewall in total at one point [on all nodes], but still no success.

    Any suggestions what to debug, how to enable the connection?






    ------------------------------
    Vid Romac
    ------------------------------


  • 2.  RE: ACE instalation on a K8s cluster - operator setup

    Posted Tue January 17, 2023 05:03 AM
    On further analysis, these changes were made

    - the OLM operator framework that was installed was updated from 1.2.0 to 1.26.0
    - the CNI kube-dns service had an invalid label selector, which caused the service not to be properly linked to the coredns pods

    also, as it was still behaving oddly
    - the coredns deployment was toggled to use `dnsPolicy` -  'ClusterFirst' then back 'Default'

    after which the subscription CRD apply worked and an install plan was generated and completed

    ------------------------------
    Vid Romac
    ------------------------------