The single_image cfmode setting is available starting with Data ONTAP 7.1. In single_image mode, a cluster has a single global WWNN, and both systems in the cluster function as a single Fibre Channel node. Each node in the cluster shares the partner node's LUN map information.
All LUNs in the cluster are available on all ports in the cluster by default. As a result, there are more paths to LUNs stored on the cluster, because any port on each node can provide access to both local and partner LUNs. You can specify the LUNs available on a subset of ports by defining portsets and binding them to an igroup. Any host in the igroup can access the LUNs only by connecting to the target ports in the portset.
For information about using portsets, see Making LUNs available on specific FCP target ports.
The following figure shows an example configuration with a multi-attached host. If the host accesses lun_1 through ports 4a, 4b, 5a, or 5b on Filer X, then Filer X recognizes that lun_1 is a local LUN. If the host accesses lun_1 through any of the ports on Filer Y, lun_1 is recognized as a partner LUN and Filer Y sends the SCSI requests to Filer X over the cluster interconnect.
Each node in the cluster shares its partner's igroup and LUN mapping information. Data ONTAP uses the cluster interconnect to check igroup and LUN mapping information and also provides the mechanisms for avoiding mapping conflicts.
When you add an initiator WWPN to an igroup, Data ONTAP verifies that there are no igroup ostype conflicts. An example ostype conflict occurs when an initiator with the WWPN 10:00:00:00:c9:2b:cc:39 is a member of an AIX igroup on one node in the cluster and the same WWPN is also a member of an group with the default ostype on the partner.
The LUN ID range on each storage system is divided into three areas:
By reserving LUN ID ranges on each storage system, Data ONTAP provides a mechanism for avoiding mapping conflicts. If the cluster interconnect is down, and you try to map a LUN to a specific ID, the lun map command fails. If you do not specify an ID in the lun map command, Data ONTAP automatically assigns one from a reserved range.
The lun online command fails when the cluster interconnect is down to avoid possible LUN mapping conflicts.
When the cluster interconnect is down, Data ONTAP cannot check for LUN mapping or igroup ostype conflicts. To avoid mapping conflicts, the following commands fail unless you use the -f option to force these commands:
You might want to override possible mapping conflicts in disaster recovery situations or situations in which the partner in the cluster cannot be reached and you want to regain access to LUNs. For example, the following command maps a LUN to an AIX igroup and assigns a LUN ID of 5, regardless of any possible mapping conflicts:
lun map -f /vol/vol2/qtree1/lun3 aix_host5_group2 5
Multipathing software is required on the host so that SCSI commands fail over to alternate paths when links go down because of switch failures or cluster failovers. In the event of a failover, none of the adapters on the takeover storage system assume the WWPNs of the failed storage system.
The following fcp config output shows how Data ONTAP displays target ports when the cluster is in single_image mode and in normal operation. Each system has two adapters. Note that all ports show the same WWNN (node name), and the mediatype of all adapter ports is set to auto. This means that the ports log into the fabric using point-to-point mode. If point-to-point mode fails, then the ports try to log into the fabric in loop mode. You can use the fcp config mediatype command to change the default mediatype of the ports from auto to another mode according to the requirements of your configuration.
Copyright © 1994-2005, Network Appliance, Inc. All rights
reserved.
Part No. 210-01094_B0
File: c7aclus4.htm
Last updated: 08-Dec-2005 13:25:31