Troubleshoot instance issues

Use this guide to diagnose common problems with CubeCOS instances (virtual machines). If you haven't already, start with the Troubleshooting overview to rule out a cluster-wide issue.

Before you begin

Some checks require the OpenStack CLI. Run these commands from a controller node, or from any host with /etc/admin-openrc.sh sourced and the openstack client installed.

An instance fails to launch

1. Check the instance status and failure reason.

   Check the instance status and fault reason

   openstack server show <instance-id> -c status -c fault

2. Open Compute > Instances, select the instance, and review the Action Logs tab for the failed event.

3. Check the following common causes:

   • The project's compute or volume quota is exhausted.
   • The source image or bootable volume isn't in Active status.
   • No host in the availability zone has enough free vCPU, memory, or disk to fit the flavor.
   • The selected network or security group no longer exists.

An instance is stuck in Build status

1. Confirm that the Compute service is healthy.

   Check cluster service health

   controller> cluster check_repair

   Look for a FIXING status on the Compute row.

2. If you launched the instance from an image, confirm that the image finished uploading and shows Active status under Compute > Images. An image stuck in Saving status with a file size of 0 usually means two identical uploads ran at the same time.

3. Review the Action Logs tab on the instance for the specific scheduling or provisioning error.

An instance won't stop or shows the wrong status

Check the server state and the compute host it's scheduled on.

Check server state and host

openstack server show <instance-id> -c status -c vm_state -c OS-EXT-SRV-ATTR:host

If status and vm_state disagree, or the reported host is unreachable, run cluster check_repair and check the Compute row before retrying the stop or start action.

The console doesn't load

1. Confirm that the instance status is Active. A console isn't available for a stopped or errored instance.
2. Reopen the console: navigate to Compute > Instances, click CONSOLE for the instance, then Confirm. For the full procedure, see Connecting to your Linux or Windows instance.
3. If the console still fails to open, run cluster check_repair and check the Compute row for a FIXING status.

Resizing or migrating an instance fails

1. Confirm the instance status is Active before you start. Resize and live migration both require the instance to be running normally.
2. After changing the flavor, confirm that you completed the Confirm Resize or Migrate step from the More > Configuration Update menu. A resize that isn't confirmed stays in a pending state.
3. Confirm that a destination host has enough free capacity for the new flavor.
4. Run cluster check_repair and check the Compute row for a FIXING status.

For the full procedure, see Resize Instance.

An instance has no network connectivity

Network connectivity problems are usually caused by the network configuration rather than the instance itself. See Troubleshoot network issues.

A volume won't attach to or detach from an instance

Volume attachment problems are usually caused by the volume or storage cluster state rather than the instance itself. See Troubleshoot storage issues.