如何扩展OpenStack Hypervisor本地磁盘容量？DevStack存储扩容咨询

Hey there! Let me walk you through the steps to add your 2TB RAID drive and expand the local storage for your DevStack-deployed OpenStack Hypervisor. I'll break this down into two main parts: preparing the RAID disk, then expanding the storage for your compute node.

1. Prepare the RAID Disk First #

Before touching OpenStack, make sure your RAID array is properly recognized and configured on the host system:

• Verify the RAID disk is detected: Run lsblk to list all block devices—you should see your 2TB RAID device (usually something like /dev/md0 or a dedicated disk name if using hardware RAID).
• Create a file system: Format the RAID device with a compatible FS (XFS is recommended for OpenStack):

  mkfs.xfs /dev/your_raid_device
  

• Get the disk UUID: Use blkid /dev/your_raid_device to grab the UUID—this is needed for persistent mounting.
• Mount temporarily (for testing):

  mkdir /mnt/openstack_new_storage
  mount /dev/your_raid_device /mnt/openstack_new_storage
  

2. Expand Hypervisor Local Storage #

How you expand depends on what you're using the local storage for (Nova instance storage or Cinder local volumes). Let's cover both common cases.

Case 1: Expanding Nova Instance Storage (Default DevStack Setup) #

DevStack defaults to storing Nova instances in /var/lib/nova/instances. Here are two safe ways to expand this:

Option A: Migrate Instance Data to the New RAID Disk

This is straightforward if you want to move all instance storage to the new drive:

• Pause/stop running instances: To avoid data corruption, pause or shut down all VMs:

  # Pause a single instance
  openstack server pause <instance_name_or_id>
  # Or shut down if downtime is acceptable
  openstack server stop <instance_name_or_id>
  

• Stop the Nova Compute service:

  systemctl stop devstack@n-compute.service
  

• Copy existing instance data: Use rsync to safely transfer data to the new mount point:

  rsync -avz /var/lib/nova/instances/* /mnt/openstack_new_storage/
  

• Backup the original directory:

  mv /var/lib/nova/instances /var/lib/nova/instances.bak
  

• Set up persistent mount:
  1. Create the new mount point: mkdir /var/lib/nova/instances
  2. Edit /etc/fstab and add this line (replace YOUR_UUID with the value from blkid):

     UUID=YOUR_UUID /var/lib/nova/instances xfs defaults 0 0
     

  3. Mount the drive: mount /var/lib/nova/instances
• Fix permissions: Ensure the Nova user owns the directory:

  chown nova:nova /var/lib/nova/instances
  

• Restart services and resume instances:

  systemctl start devstack@n-compute.service
  # Resume instances
  openstack server unpause <instance_name_or_id>
  # Or start if you shut them down
  openstack server start <instance_name_or_id>
  

Option B: Expand with LVM (If Instance Storage is on a Logical Volume)

If your /var/lib/nova/instances is already on an LVM volume, you can extend it with the new RAID disk:

• Add the RAID disk to your existing volume group:

  vgextend <your_volume_group_name> /dev/your_raid_device
  

• Expand the logical volume:

  lvextend -l +100%FREE /path/to/your/logical_volume
  

• Resize the file system:
  • For XFS: xfs_growfs /path/to/your/logical_volume
  • For EXT4: resize2fs /path/to/your/logical_volume

Case 2: Expanding Cinder Local LVM Storage #

If you're using Cinder's local LVM backend (DevStack enables this by default), adding the RAID disk to the Cinder volume group is simple:

• Add the RAID disk to the cinder-volumes volume group:

  vgextend cinder-volumes /dev/your_raid_device
  

• Cinder will automatically detect the new free space—you don't need to restart any services. New volumes you create will use the expanded storage.

Key Notes to Remember #

• Backup first: Always back up critical data (instance images, Cinder volumes) before making storage changes.
• RAID validation: Double-check your RAID array's health (use tools like mdadm for software RAID or your hardware RAID manager) before proceeding.
• Test in staging: If this is a production-like environment, test these steps in a staging setup first to avoid surprises.

内容的提问来源于stack exchange，提问作者Inmd