Taking over a UVM span
A span which resides on external storage and is accessed by one cluster can be taken over by another cluster. For example, the span is on a storage system which is being accessed by Cluster A and you now want it to be accessed by Cluster B.
The following procedure also ensures that the two clusters never access the same span at the same time. This prohibition is necessary for data integrity.
To take over a span using UVM:
Procedure
On Cluster A:
Cleanly unmount all of the file systems.
Run the span-assign-to-cluster <span-name> <UUID> <cluster-B-name> command, specifying the name and UUID of Cluster B.
Run the span-deny-access <span-name> command.
Use LUN-mapping, switch-zoning or both to ensure that Cluster A can no longer access the LUs in the span.
Following the instructions in the Hitachi Universal Volume Manager User Guide, set up UVM internal LUs on Cluster B's local storage and make them visible to Cluster B by assigning host LUNs to them.
On Cluster B:
Depending on the configuration of your storage, it may be necessary to run the scsi-refresh command on the server to enable it to see the new internal LUs.
Run the sd-list --hdp all:denied command to identify the new UVM internal LUs.
Use the sd-allow-access command to license the UVM internal LUs. If there are no other unlicensed SDs, you can license the internal LUs more easily by running the sd-allow-access all:denied command.
Run the span-wait-for-loading command to give the span and filesystem time to load their disk configuration.
If the span-list command does not show the span, use the sd-recover-from-changed-luids --span-base-name and span-rewrite-cod commands to recover the span.
Use the evsfs command to bind each file system to an EVS.
Mount each file system and create the necessary exports and shares.