Ensure that you are familiar with the constraints for zoning
a switch to support the Metro Mirror and Global Mirror feature.
SAN configurations that use intracluster Metro Mirror
and Global Mirror relationships do not require additional switch zones.
SAN
configurations that use intercluster Metro Mirror and Global Mirror
relationships require the following additional switch zoning considerations:
- A cluster can be configured so
that it can detect all nodes in all remote clusters. Alternatively,
a cluster can be configured so that it detects only a subset of nodes
in remote clusters. For I/O groups that share an intercluster Metro
Mirror or Global Mirror relationship, all node ports in each I/O group must be zoned
to detect all node ports in the other I/O group.
- Use of interswitch link (ISL) trunking in a switched fabric.
- Use of redundant fabrics.
For intercluster Metro Mirror and Global Mirror relationships,
you must perform the following steps to create the additional zones
that are required:
- Configure your SAN so that fibre-channel traffic can be passed
between the two clusters. To configure the SAN this way, you can connect
the clusters to the same SAN, merge the SANs, or use routing technologies.
- (Optional) Configure
zoning to allow all nodes in the local fabric to communicate with
all nodes in the remote fabric.
Note: - If you are using McData Eclipse routers, model 1620, only 64 port
pairs are supported, regardless of the number of iFCP links that are
used.
- (Optional) As an alternative to
step 2,
choose a subset of nodes in the local cluster to be zoned to the nodes
in the remote cluster. Minimally, you must ensure that one whole I/O
group in the local cluster has connectivity to one whole I/O group
in the remote cluster. I/O between the nodes in each cluster is then
routed to find a path that is permitted by the configured zoning.
Reducing
the number of nodes that are zoned together can reduce the complexity
of the intercluster zoning and might reduce the cost of the routing
hardware that is required for large installations. Reducing the number
of nodes also means that I/O must make extra hops between the nodes
in the system, which increases the load on the intermediate nodes
and can increase the performance impact; in particular, for Metro
Mirror.
- Optionally, modify the zoning so that the hosts that are visible
to the local cluster can recognize the remote cluster. This allows
a host to examine data in both the local and remote cluster.
- Verify that cluster A cannot recognize any of the back-end storage
that is owned by cluster B. A cluster cannot access logical units (LUs) that a host
or another cluster can also access.