Learn how to fix OpenStack volume deletion failures caused by Ceph configuration issues using cinder-manage commands from the controller node.

1. Problem Overview - Understanding the issue
2. Prerequisites - What you need before starting
3. Solution Steps - The fix using cinder-manage
4. Verification - Confirm the fix worked
5. Prevention - How to avoid this issue in the future

Overview

When OpenStack volumes fail to delete, I ran into a situation where the 'host' set on the volume was referencing an incorrect backend host, and I was unable to delete teh volumes, even forcibly. This is based around a Kolla-Ansible deployment. This typically happens when:

• There is a misconfiguration or pools are reconfigured or renamed
• Storage backends are changed
• Host information becomes inconsistent between Cinder and Ceph
• Volume host references point to non-existent or misconfigured storage locations

The solution involves using cinder-manage from within the cinder-volume container on the controller node to update the volume host information.

1. PROBLEM OVERVIEW

Symptoms:

• OpenStack volume deletion commands fail with errors like:

  ERROR: Volume deletion failed
  

• Volumes appear stuck in "deleting" state
• Cinder logs show host-related errors

Root Cause: The volume's host information in the Cinder database doesn't match the actual Ceph configuration, preventing proper volume deletion.

2. PREREQUISITES

Before attempting this fix, ensure you have:

• Controller Node Access: SSH access to your OpenStack controller node
• Admin Privileges: OpenStack admin credentials
• Volume Information: The volume ID that's failing to delete
• Ceph Pool Details: Current Ceph pool configuration information

Required Commands:

# Check current volume status
openstack volume show <volume-id>

# Verify Ceph pool configuration
ceph osd pool ls
ceph osd pool ls detail

3. SOLUTION STEPS

Follow these steps to fix the volume deletion issue:

Step 1: Access the Cinder Volume Container

From your controller node, access the cinder-volume container:

# List running containers to find cinder-volume
docker ps | grep cinder-volume

# Access the cinder-volume container
docker exec -it <cinder-volume-container-id> bash

Step 2: Identify Current Host Information

Inside the container, check the current host information for the problematic volume:

# Check volume details in Cinder database
cinder-manage volume show <volume-id>

Look for the host field in the output, which will show something like:

host: controller03@rbd-2#rbd-2

Step 3: Update Volume Host Information

Use cinder-manage to update the volume's host information to match your current Ceph configuration:

# Update the volume host information
cinder-manage volume update_host \
  --currenthost controller03@rbd-2#rbd-2 \
  --newhost rbd:volumes_tier2@rbd-2#rbd-2

Command Breakdown:

• --currenthost: The current (incorrect) host information
• --newhost: The correct host information matching your Ceph configuration
• Format: backend:pool@ceph-cluster#ceph-cluster

Step 4: Exit Container

exit

4. VERIFICATION

After updating the host information, verify the fix:

Test Volume Deletion

# Attempt to delete the volume
openstack volume delete <volume-id>

# Check volume status
openstack volume show <volume-id>

Verify Database Update

You can also verify the change was applied correctly:

# Access the container again
docker exec -it <cinder-volume-container-id> bash

# Check the updated volume information
cinder-manage volume show <volume-id>

5. PREVENTION

To avoid this issue in the future:

Best Practices

1. Document Configuration Changes: Keep detailed records of Ceph pool changes
2. Test in Staging: Always test storage configuration changes in a staging environment first
3. Monitor Volume Operations: Set up monitoring for volume deletion failures
4. Regular Audits: Periodically audit volume host information consistency

Configuration Management

# Regular health check script
#!/bin/bash
# Check for volumes with inconsistent host information
cinder-manage volume list | grep -E "controller[0-9]+@rbd-[0-9]+#rbd-[0-9]+"

COMMON SCENARIOS

Scenario 1: Pool Rename

If you renamed a Ceph pool from volumes to volumes_tier2:

cinder-manage volume update_host \
  --currenthost controller03@volumes@rbd-2#rbd-2 \
  --newhost controller03@volumes_tier2@rbd-2#rbd-2

Scenario 2: Backend Change

If you changed from lvm to rbd backend:

cinder-manage volume update_host \
  --currenthost controller03@lvm#lvm \
  --newhost controller03@rbd-2#rbd-2

TROUBLESHOOTING

If the Command Fails

1. Check Volume Status: Ensure the volume isn't in use

   openstack volume show <volume-id> | grep status
   

2. Verify Host Format: Double-check the host format matches your Ceph configuration

   ceph osd pool ls detail
   

3. Check Cinder Logs: Review logs for additional error details

   docker logs <cinder-volume-container-id>
   

Alternative Approaches

If cinder-manage doesn't work, you can also:

1. Direct Database Update: Update the Cinder database directly (use with caution)
2. Volume Migration: Migrate the volume to a working backend
3. Force Delete: As a last resort, force delete from Ceph directly

SUMMARY

This guide walked you through fixing OpenStack volume deletion issues caused by Ceph configuration problems. The key steps were:

1. Identify the problematic volume and its current host information
2. Access the cinder-volume container on the controller node
3. Update the volume host using cinder-manage volume update_host
4. Verify the fix by attempting volume deletion
5. Prevent future issues through proper configuration management

The cinder-manage volume update_host command is a powerful tool for resolving host-related volume issues in OpenStack environments with Ceph storage backends.

Next Steps

After fixing the volume deletion issue, consider:

• Implementing monitoring for similar issues
• Documenting your Ceph configuration changes
• Setting up automated health checks
• Reviewing your storage configuration management processes

For more information about OpenStack Cinder management, visit the official OpenStack documentation.

Need help with your OpenStack deployment? Get $200 in free credits and start hosting your applications on Gozunga Cloud today!