Managing nodes in a storage system

1. A method of managing changes in ownership of virtual volume regions between nodes in a storage system, comprising:

creating zero logical disk (LD) groups, each zero LD group comprising zero LDs from different sets of zero LDs owned by different nodes, each zero LD having a unique pair of a primary node and a backup node;

creating thinly-provisioned virtual volumes (TPVVs) by mapping each TPVV to one zero LD group; and

when a change of ownership occurs, switching ownership of affected zero LDs one at a time so host IO is suspended for one group of the TPVVs at a time.

2. The method of claim 1 , wherein each TPVV comprises virtual volume regions mapped to zero LDs in a corresponding zero LD group.

3. The method of claim 1 , wherein each TPVV is mapped to a zero LD group different from the most recently used zero LD group.

4. The method of claim 1 , wherein each zero LD is associated with a primary node, a backup node, and a group ID.

5. The method of claim 1 , wherein switching ownership of the affected zero LDs one at a time comprises:

determining TPVVs mapped to a zero LD group with an affected zero LD;

suspending host IOs the TPVVs mapped to the zero LD group; and

flushing data from cache to backing store for an old primary node.

6. The method of claim 1 , further comprising:

when the change of ownership occurs, switching ownership of affected non-zero LDs from one common provisioning group at a time so host IO is suspended for one group of the common-provisioned virtual volumes (CPVVs) at a time, the CPVVs being fully provisioned, the common provisioning group providing a pool of LD regions for creating virtual volumes.

7. The method of claim 6 , wherein switching the ownership of the affected non-zero LDs comprises:

determining non-zero LDs having an old primary node;

determining CPVVs mapped to the non-zero LDs having the old primary node;

suspending host IO to the group of the CPVVs mapped to the non-zero LDs having the old primary node; and

flushing data from cache to backing store for the old primary node.

8. The method of claim 1 , wherein the change of ownership occurs when a node joins the storage system or when characteristics of a VV is changed.

9. A method of managing nodes in a storage system, comprising:

determining thinly-provisioned virtual volumes (TPVVs) mapped to different zero logical disk (LD) groups, each zero LD group comprising different zero LDs owned by different nodes;

suspending host IOs to a group of TPVVs mapped to one zero LD;

determining non-zero LDs owned by an old primary node;

determining common-provisioned virtual volumes (CPVVs) mapped to the non-zero LDs owned by the old primary node, the CPVVs being fully provisioned;

suspending host IO to the CPVVs mapped to the non-zero LDs owned by the old primary node; and

flushing data from cache to backing store for the old primary node.

10. The method of claim 9 , further comprising:

creating the zero LD groups, wherein each zero LD in a zero LD group has a unique pair of a primary node and a backup node;

creating the TPVVs by mapping each TPVV to one zero LD group; and

creating the CPVVs by mapping each CPVV to non-zero LDs from a common provisioning group, the common providing group providing a pool of LD regions for creating virtual volumes.

11. A storage system comprises:

a node comprising:

zero logical disk (LD) groups each comprising zero LDs, the zero LDs having a unique pair of primary and backup nodes within its zero LD group;

thinly-provisioned virtual volumes (TPVVs) each mapped to one zero LD group; and

a system manager to create the zero LD groups, to create the TPVVs, and, when a change of ownership occurs, switch ownership of affected non-zero LDs one at a time so host IO is suspended for one group of the TPVVs at a time.

12. The system of claim 11 , wherein the system manager to map each

TPVV to a zero LD group different from the most recently used zero LD group.

13. The system of claim 11 , wherein the system manager to switch ownership of the affected zero LDs one at a time comprises:

determine TPVVs mapped to an affected zero LD;

suspend host IOs the TPVVs mapped to the affected zero LD; and

flush data from cache to backing store for an old primary node.

14. The system of claim 11 , wherein:

the node further comprising:

common-provisioned virtual volumes (CPVVs) each mapped to at least one non-zero LD from a common provisioning group, the CPVVs being fully provisioned, the common provisioning group providing a pool of LD regions for creating virtual volumes; and

when the change of ownership occurs, to switch ownership of the non-zero LDs one common provisioning group at a time so host IO is suspended for one group of the CPVVs at a time.

15. The system of claim 14 , wherein to switch ownership of the non-zero LDs comprises:

to determine non-zero LDs having an old primary node;

to determine CPVVs mapped to the non-zero LDs having the old primary node;

to suspend host IO to the CPVVs mapped to the non-zero LDs having the old primary node; and

to flush data from cache to backing store for the old primary node.