Troubleshooting cluster creation

The cluster hangs in the Provisioning state

Error creating VM

The expected sequence of creating nodes from a template and the connection of templates with the cluster are described on the page Creating nodes when deploying a client cluster.

  • Check the VM template and its availability;
  • Check your provider account permissions.

See section “Troubleshooting providers”.

Nodes do not go to Ready

  • Connect to the node via SSH and check the status of kubelet, containerd;
  • Check the network connection between the nodes and to the API server;
  • View bootstrap provider logs.

There are no free IPs in the pool

  • Add addresses to pool or create new pool;
  • Release unused addresses from remote clusters.

VIP for API server or Ingress is already taken

When creating a cluster, the platform checks the availability of the addresses you specified for KubeAPI and Ingress. If the address is already in use on the network, the deployment may fail during the cluster creation phase.

How to check

Perform the check from the same network zone from which the platform deploys the client cluster: connect via SSH to the management cluster node running under shturval-backend and check the required ports.

      Checking the port using nc:

      nc <IP-ADDRESS-KUBEAPI> 6443 -v -w 5

      Checking HTTPS endpoint using curl:

      curl -k https://<IP-ADDRESS-KUBEAPI>:6443

        Checking the port using nc:

        nc <VIP-ADDRESS-INGRESS> 443 -v -w 5

        Checking HTTPS endpoint using curl:

        curl -k https://<VIP-ADDRESS-INGRESS>:443

        If the response contains Connection refused or Forbidden, the address is already responding on the network and cannot be used as a free VIP for the new cluster. Free the address, select another VIP or expand the IP pool, then recreate the cluster.

        Error due to network or subnet crossing

        ← How to deploy a cluster
        Checking the intersection of subnets when creating a cluster →