This task provides step-by-step instructions about how to add nodes to the cluster using the SAN Volume Controller Console. A cluster must have a minimum of two nodes, otherwise, you must add at least one other node to the cluster so it has a minimum of two nodes.
| Attention: Before adding a node to a cluster, make sure that the switch zoning is configured such that the node being added is in the same zone as all other nodes in the cluster. In particular, if you are replacing a node and the switch is zoned by worldwide port name (WWPN) rather than by switch port, you must update the switch configuration. |
Before adding a node to the cluster check to see if any of the following conditions are true:
| Attention: If any of these conditions are true, then you must perform the following special procedures. Failure to perform the special procedure is likely to result in the corruption of all data managed by the cluster. |
If any of the previous conditions are true, then the following special procedures apply. These special procedures apply when you use either the svctask addnode command or the SAN Volume Controller Console. When a node is added to a cluster then either:
Note: The WWNN of the nodes in the cluster can be determined
using the command:
svcinfo lsnode |
The following are two hypothetical scenarios where the special procedures may apply:
Applications on host systems direct I/O operations to filesystems or logical volumes which are mapped by the operating system to vpaths which are pseudo disk objects supported by the SDD driver. See the IBM TotalStorage Subsystem Device Driver: User's Guide.
The SDD driver maintains an association between a vpath and a SAN Volume Controller VDisk. This association uses an identifier (UID) which is unique to the VDisk and is never re-used. This allows the SDD driver to unambiguously associate vpaths with VDisks.
The SDD device driver operates within a protocol stack which also contains Disk and Fibre Channel device drivers which allow it to communicate with the SAN Volume Controller using the SCSI protocol over Fibre Channel as defined by the ANSI FCS standard. The addressing scheme provided by these SCSI and Fibre channel device drivers uses a combination of a SCSI Logical unit number (LUN) and the World Wide Name for the Fibre Channel Node and Ports.
In the event of errors occurring, error recovery procedures (ERPs) operate at various tiers in the protocol stack. Some of these ERPs cause I/O to be redriven using the same WWN and LUN numbers which were previously used.
The SDD device driver does not check the association of the VDisk with the VPath on every I/O that it performs.
Consider a four-node SAN Volume Controller configuration.
The nodes, Node1 and Node2, are in I/O group 0 which supports the VDisk, VDisk0.
The nodes, Node3 and Node4, are in I/O group 1 which supports the VDisk, VDisk1.
Assume that VDisk 0 is mapped to a host as LUN 0. This will be LUN 0 associated with the ports in Node1 and Node2. We might represent this as N1/0 and N2/0 respectively. Assume also that VDisk1 is also mapped to the host as LUN 0. Thus N3/0 and N4/0 are mapped to VDisk1.
Now assume that nodes, Node2 and Node4, are removed from the cluster.
If Node2 is added back into the cluster into I/O Group 1 a data corruption could occur because:
Assume that the cluster has been created.
Perform the following steps to add nodes to the cluster:


| Attention: If you are re-adding a node to the SAN, ensure
that you are adding the node to the same I/O group that it was removed from.
Failure to do this can result in data corruption. Use the information that
was recorded when the node was originally added to the cluster. If you do
not have access to this information, call IBM Service to add the node back
into the cluster without corrupting the data.
This can avoid a possible data corruption exposure if the node must be removed from and re-added to the cluster.
|

Notes:
|
In our hypothetical scenario, the nodes are called:
knode and lnode
In our hypothetical scenario, the I/O group is called:
io_group0
In our hypothetical scenario, the nodes are called:
mnode and nnode
In our hypothetical scenario, the I/O group is called:
io_group1