NFS cluster-specific issues
•Back up user and group mappings.
To avoid loss of complex advanced mappings in the case of a system failure, back up the mappings whenever the mappings have been edited or new mappings have been added.
•Map consistently.
Groups that are mapped to each other should contain the same users and the members of the groups should be properly mapped to each other to ensure proper file access.
•Map properly.
•Valid UNIX users should be mapped to valid Windows users.
•Valid UNIX groups should be mapped to valid Windows groups.
•Mapped Windows users must have the “Access this computer from the Network privilege” or the mapping will not function properly.
•Mapped Windows users must have an active password, or the mapping will not function properly.
•In a clustered deployment, create user name mappings using domain user accounts.
Because the security identifiers of local accounts are recognized only by the local server, other nodes in the cluster will not be able to resolve those accounts during a failover. Do not create mappings using local user and group accounts.
•In a clustered deployment, administer user name mapping on a computer that belongs to a trusted domain.
If NFS administration tasks are performed on a computer that belongs to a domain that is not trusted by the domain of the cluster, the changes are not properly replicated among the nodes in the cluster.
•In a clustered deployment, if PCNFS password and group files are being used to provide user and group information, these files must be located on each node of the system.
Example: If the password and group files are located at c:\maps on node 1, then they must also be at c:\maps on node 2. The contents of the password and group files must be the same on both nodes as well.
These password and group files on each server node must be updated periodically to maintain consistency and prevent users or groups from being inadvertently squashed.
Non-cluster aware file sharing protocols
Services for Macintosh (SFM), File and Print Services for NetWare, HTTP file sharing protocols are not cluster aware and will experience service interruption if installed on a clustered resource during failover events of the resource. Service interruptions will be similar to those experienced during a server outage. Data that has not been saved to disk prior to the outage will experience data loss. In the case of SFM, it is not supported because SFM maintains state information in memory. Specifically, the Macintosh volume index is located in paged pool memory. Using SFM in clustered mode is not supported and may result in data loss similar in nature to a downed server should the resource it is based on fails over to the opposing node.
Adding new storage to a cluster
Present the new storage to one node in the cluster. This can be accomplished through selective storage presentation or through SAN zoning.
P4000 G2 Unified NAS Gateway User Guide 45
