The ABI cluster installation appeared successful because the agent wait-for install-complete command only checks if all operators are available, but in reality, the etcd operator was still in progress.

I attempted to debug the issue and discovered that the etcd pod on the rendezvous node was in a CrashLoopBackOff state.

        message: |
          /bin/sh: line 4: NODE_ci_ln_8hx21zb_c1627_master_0_ETCD_URL_HOST: not set
        reason: Error
        startedAt: "2024-10-15T16:25:07Z"
    name: etcd-ensure-env-vars
    ready: false
    restartCount: 6
    started: false
    state:
      waiting:
        message: back-off 5m0s restarting failed container=etcd-ensure-env-vars pod=etcd-ci-ln-8hx21zb-c1627-master-0_openshift-etcd(4e70d2cb4965143304a7aaf1d7e91a5e)
        reason: CrashLoopBackOff 

4.18.0-0.nightly-2024-10-15-032107

Always

ClusterOperators:664 clusteroperator/etcd is progressing: NodeInstallerProgressing: 1 node is at revision 0; 1 node is at revision 5; 1 node is at revision 6

Installation should be successful.