SA Blog --系统管理员的博客生涯

Changing the Host Mode setting on the HDS storage array leads to all of the VMFS3 volumes being seen as snapshot volumes by the ESX Server system. This results in the volumes being disabled.

 

Similar behavior is observed when setting the SPC-2 flag on the EMC Symmetrix storage array. How is this issue resolved?

Making this type of change on the storage array results in assigning a new LUN ID (UUID or Serial Number) to all LUNs presented via the modified ports/FAs.

VMFS3 metadata identifies the volumes by several properties which include the LUN number and the LUN ID (UUID or Serial Number). Because the LUNs now have new UUIDs, the resulting mismatch with the metadata leads to LVM identifying the volumes as snapshots.

 

This procedure allows you to make all of the VMFS3 volumes visible again.

 

Important: If you do not have actual snapshot LUNs presented and have no plans to present any to these hosts, follow the directions in http://kb.vmware.com/kb/6482648, and then skip to step 9 below.

1. Stop the running virtual machines on all the ESX Server hosts.
    
2. Rescan. You see warnings about “snapshot LUN” referenced in /var/log/vmkernel.
    
3. Enable LVM Resignaturing on the first ESX Server host (set LVM.EnableResignature to 1):
    
   
   1. Log on to the ESX Server host with VI Client.
       
   2. Click the Configuration tab.
       
   3. Select the Advanced setting option.
       
   4. Choose the LVM section.
       
   5. Set the value of LVM.EnableResignaturing to 1.
       
   6. Save the change.
       
   7. Click the storage adapter.
       
   8. Click Rescan Adapter.
       
   9. Leave the default option and proceed.

You are now be able to see the VMFS volumes with labels prefixed with snap. 

4. Disable LVM Resignaturing (set LVM.EnableResignature to 0):
   
   
   1. Log on to the ESX Server host with VI client.
       
   2. Click the Configuration tab.
       
   3. Select the Advanced setting option.
       
   4. Choose the LVM section.
       
   5. Set the value of LVM.EnableResignaturing to 0.
       
   6. Save the change.

No snapshot messages are visible in /var/log/vmkernel.

5. Re-label the volume.
    
   
   1. Log into VirtualCenter Server with VI Client. (Connecting directly to the ESX Server host does not offer the choice identified in step b).
       
   2. In Inventory view, select Datastores view.
       
   3. Select the datastore. Right click and select Remove to remove the old label, which is associated with the old UUID of the volume.
       
   4. Select Hosts & Clusters view.
       
   5. In the Summary tab, you see the list of datastores.
       
   6. Click in the name field for the volume and change snap0000 to the original name.

You now have the correct original label associated with the resignatured volume.

6. Rescan storage from all ESX Server hosts. You do not need to rescan from the host that performed the resignature.
    
7. Re-register all the virtual machines. Because the virtual machines are registered against the old UUID, you must re-register them in VirtualCenter.
    
   1. Log on to the ESX Server host with VI client.
       
   2. Click the Configuration tab.
       
   3. Select Storage (SCSI, SAN & NFS).
       
   4. Double-click any of the datastores to open the Datatstore browser.
       
   5. Navigate to the .vmx file of any of the virtual machines by clicking the folders.
       
   6. Right click and select Add to inventory.
       
8. Remap any RDMs.

If you have a virtual machine that uses an RDM, you need to recreate the mapping.

Caution: If you used multiple RDMs, it may be difficult to identify one from another. If the RDMs are different sizes, you can map them in the correct order by their sizes. If the RDMs are all the same size, this is a more difficult process because you must map the RDMs to one virtual machine at a time, boot the virtual machine, and then verify if it is the correct LUN. To prevent having to map the RDMs to one virtual machine at a time:

9. Power on the virtual machines. Reply yes if prompted about a new UUID.

If any of the virtual machines refer to missing disks when they power up, check the.vmx file and ensure that the SCSI disk references are not made against the old UUID instead of against the label (or new label, if you changed it).

10. Repeat steps 3 through 9 for all subsequent ESX Server hosts that still see snapshot volumes. If all ESX Server hosts share the same volumes, this step is not necessary.