The partner cfmode is the default setting for all new systems. It is supported on all FCP-licensed systems except for the FAS270c. It is also supported for all host OS types.
For systems with HBAs, Port A and Port B are both active. Port A on each HBA provides access to local LUNs, and Port B provides access to LUNs on the partner system. The target ports log into the fabric using a point-to-point topology.
If you have a FAS3000 series system with a new installation of DATA ONTAP, the state of the onboard Fibre Channel port depends on your configuration. In the default two-port configuration, ports 0c and 0d connect to the SAN. Port 0c provides access to local LUNs, and port 0d provides access to LUNs on the partner. In a four-port configuration in which all onboard ports connect to the SAN, port 0a and 0c on each node in the cluster provide access to local LUNs, and ports 0b and 0d provide access to LUNs on the partner.
The following figure shows a sample configuration with a multi-attached host connecting to a cluster with target HBAs. The solid lines represent paths to LUNs on the local storage system, the dotted lines represent paths to partner LUNs.
Partner mode requires hostside multipathing software. The multipathing policy is active/passive. The primary paths to the LUNs are always through the A ports. The B ports are secondary paths.
The following table shows the available paths between the host and the storage system cluster in the preceding configuration example.
When a link fails--for example, Switch 1 in the preceding example fails--the host loses a primary path (through the A port of the HBA in Slot M) to Filer X. The host fails over to the other primary paths. If there are no other primary paths available, the host can access LUNs through the secondary paths (B ports) on Filer Y.
The failover method depends on the host and multipathing software. For example, if you have VERITAS® Volume Manager (VxVM) with Dynamic Multipathing software and the Array Support Library (ASL) on a Solaris host, all LUNs that share active paths form a group. If all active paths fail for a LUN in a group, all LUNS in the group fail over to the secondary paths. For detailed information about how each host handles failover, see the ASL documentation for your FCP host attach kit.
If Filer Y takes over for Filer X, the host continues to access LUNs on Filer X through the B ports on Filer Y. The WWNN and WWPN of the B ports on Filer Y do not change. This enables HP-UX and AIX hosts, which track target devices based on WWPN/WWNN and N_Port ID (the switch-assigned addresses), to maintain correct information about available paths.
When the FCP cfmode setting is partner, the local and partner addresses of the WWNN and WWPN have a pattern of 50:a9:80:nn:nn:nn:nn. The WWPN and WWNN of the B ports are based on the WWNN of the partner storage system in the cluster. For example, port B on Filer X presents the WWNN of Filer Y. This means that the WWNNs and WWPNs do not change during a takeover. The static address requirements for the N_Port ID, S_ID and D_ID are maintained.
The following fcp config command output shows how Data ONTAP displays WWNN and WWPN when the storage system's cfmode is set to partner and the cluster is in normal operation.
Copyright © 1994-2005, Network Appliance, Inc. All rights
reserved.
Part No. 210-01094_B0
File: c7aclus3.htm
Last updated: 08-Dec-2005 13:25:31