Rules For Operating a NetWare Servers on a SAN with or without Clustering Services

Rules For Operating a NetWare Servers on a SAN with or without Clustering Services

When you create a NCS Storage Area Network, it is important to remember that all servers attached to the shared device, whether in the cluster or not, have access to all the volumes on the shared storage unless you specifically prevent such access.

NCS arbitrates access to shared volumes for all cluster nodes, but cannot protect shared volumes from being corrupted by non-cluster servers.

When working with shared storage, you must observe these rules:

• You must not install NetWare on a server that is attached to a SAN.The NetWare install will delete all NW partitions that it finds. If it is attached to shared storage, it will delete those as well. You must disconnect servers from their FC cables before installing NetWare. This is true of virtually any OS install--NT's install will behave exactly the same way.
  
• You must not have a non-cluster server attached to the same shared storage device as a cluster, unless you have zoned off the storage such that the non-cluster server accesses only its own volumes, and not cluster volumes. Otherwise it could result in data corruption.
• Prior to creating or modifying partitions on a SAN ensure all servers see the same partition table. This can be done by issuing a MM SCAN FOR NEW PARTITIONS command prior to any disk carving, and this will ensure that the server does not overwrite partitioned data that may have been created from another server. (This is needed because the server will only read the partition table when it is booted or forced by the MM SCAN FOR NEW PARTITIONS )
• If working with a new SAN and wishing to set up all the pools and volumes prior to setting up Cluster Services, it would be best if all the partitions/pools were created from one server, to ensure that all the partition information is correct. This being done, the NSS pools can be mounted via NSSMU or command line NSS /POOLACTIVATE=poolname, then MOUNT volumename to give the server access. IMPORTANT - ONLY ONE SERVER CAN ACTIVATE A POOL AT ONE TIME. No 2 servers can mount the same volume at the same time. NSS will prevent this unless the default MSAP setting was changed (NOT RECOMMENDED). To finalize the pool/volume creation update NDS/eDirectory from the server that will host the pools. This can be done via iManager, NRM, or NSSMU.
• Ultimately if Cluster Services is installed, this will prevent the partition overwriting other partitions. NCS will update all servers in the cluster each time a clustered enable resource is created. Thus preventing possible corruption.
• You must not perform NSS cluster volume operations (deletion, resizing, renaming, etc.) from non-cluster nodes. A cluster node is one that is running the NWCS NLMs.

Additional tips:

• How to Install NetWare on a Server that Will be Added to an Existing Cluster
• • Install your fibre channel hardware. This will cause NetWare to detect and load the proper drivers when it installs.
  • Unplug the fibre channel card from the storage. Do this by physically removing the FC cable.
  • Install NetWare.
    

• How to Add a New Node to an Existing Cluster
• • Install NetWare on the node.
  • Let NetWare boot.
  • Reconnect the server to the shared storage.
  • Run the cluster install.
  • If you have changed the default failover scripts, add this new node to the failover lists. (By default, NWCS will fail over to all existing cluster nodes, including newly added nodes.)
    

• How to Re-Add a Node to a Cluster that was Previously in the Cluster
• • Unplug the node from the SAN
  • Install NetWare on the node using the same node name and IP address
  • Using ConsoleOne, ensure that the cluster node object in the cluster container is deleted. (Go to the Cluster container. On the right frame, highlight the node to be deleted. Hit the delete key.)
  • Answer yes to delete confirmation.
  • Plug node into SAN
  • Run NWCS install. Node will take on its former identity.

• How to Run NSS Verify or Rebuild on a Cluster Volume (NetWare 5.0) or Pool (NetWare 6.x)
• • Offline the resource
  • Run Verify or Rebuild from any cluster node

The following are procedures to follow when operating a NetWare 5.1 cluster (iManager should be used when operating a NetWare 6.5 cluster):

• How To Add Free Space to an Existing Cluster-Enabled Volume
  
• • Execute a "CLUSTER DOWN" to bring down the cluster.
  • Execute a "DOWN" on every server node but one. If you have an NDS master replica in the cluster, it is best to have that node be the one left running.
  • Perform the volume operation from the remaining node.
  • Bring the other servers back up and allow them to join the cluster.

• How to Delete a Cluster-Enabled Volume
  
• • Execute a "CLUSTER DOWN" to bring down the cluster.
  • Execute a "DOWN" on every node but one. If you have an NDS master replica in the cluster, it is best to have that node be the one left running.
  • Using NSS Menu, delete the cluster volume. (You may optionally release ownership of the storage.)
  • Using ConsoleOne, delete the resource that used that volume.
  • Bring the other servers back up and allow them to join the cluster.
    

• How to Create a Cluster-Enabled Volume
  
• • Execute a "CLUSTER DOWN" to bring down the cluster.
  • Execute a "DOWN" on every node but one. If you have an NDS master replica in the cluster, it is best to have that node be the one left running.
  • Using NSS Menu, create the cluster volume.
  • Using ConsoleOne, cluster-enable the volume.
  • Bring the other servers back up and allow them to join the cluster.
    

1. For (NetWare 5.1) NCS 1.01 SP2 and later, you may perform all volume operations from any node but still needs to be done from the node in which the Volume is active and Clustering Services software is loaded and running on each node. NWCONFIG.NLM, NSS and Clustering are now more tightly integrated and should give an error message if trying to perform volume operations from a node other than the node the volume is active.
   
   
2. Using a SAN for multiple servers without clustering is NOT recommended without either LUN masking or port zoning. This is due to the possibility of data loss from either a new server install while attached to the SAN or from mounting a single volume on more than one server at the same time.

Formerly known as TID# 10057924

See also TID 3084136 - Avoiding problems with Cluster Services (https://www.novell.com/support/search.do?searchString=3084136)