Troubleshooting the global Cluster on VMware vSphere

For the installation stages and the failures every provider shares, see Troubleshooting. The following are specific to VMware vSphere. Run the commands below on the bootstrap node; unless a row says otherwise, they read the bootstrap cluster, which exists until the installation completes. If /var/cpaas/data/alauda.kubeconfig is missing, see Common Failures.

ResultAction
Preflight reports a name that was not foundCorrect the name in install-config.yaml, then run create again. Preflight stops at the first missing object, so repeat until it passes. With several compute resources in the datacenter, set computeResource to the name of one.
Preflight passes, but a VM is not created because a datastore, folder, or port group is not foundContact technical support to correct the inventory path.
Preflight cannot sign inCorrect server, username, and password in install-config.yaml, check that vCenter answers on TCP/443 from the bootstrap node, then run create again.
Stage 4 does not advanceRun <extract-dir>/res/<arch>/ac --kubeconfig /var/cpaas/data/alauda.kubeconfig -n cpaas-system describe vspheremachine <name> and <extract-dir>/res/<arch>/ac --kubeconfig /var/cpaas/data/alauda.kubeconfig -n cpaas-system describe vspherevm <name>. If the conditions name datastore capacity, free space on it. Otherwise, contact technical support.
A VM is created but never joinsFrom the VM console, check that the node reaches the control plane endpoint on TCP/6443 and the registry, with the gateway and DNS from install-config.yaml, and fix the network. If the node reaches both, contact technical support.
Nodes stay uninitialized after the platform installationBefore the handover, run <extract-dir>/res/<arch>/ac --kubeconfig /var/cpaas/data/alauda.kubeconfig -n cpaas-system get clusterresourcesetbinding global -o yaml and check that global-vsphere-cpi is applied. After the handover, check the CPI Pods on the global cluster. If neither shows a cause, contact technical support.
Stage 8 waits for the credentialCheck that cpaas-system/global-credential exists on the global cluster and is not terminating. If it is missing or stays terminating, contact technical support.
curl … /healthz through the Self-built VIP does not print okCheck that the network allows VRRP (IP protocol 112, multicast 224.0.0.18) and gratuitous ARP between the control plane nodes, and that no other VRRP instance in the Layer 2 domain uses the vrid. If both are met, contact technical support.

For provider failures that also apply after installation, see Troubleshooting VMware vSphere Workload Clusters.