Example: Two-node NFS cross-mounting configuration

The this example, Node A currently hosts a nonconcurrent resource group, RG1, which includes /fs1 as an exported NFS file system and service1 as a service IP label.

In this example, Node B currently hosts a nonconcurrent resource group, RG2, which includes /fs2 as an exported NFS file system and service2 as a service IP label. On reintegration, /fs1 is passed back to Node A, locally mounted, and exported. Node B mounts it over NFS again.

Figure 1. Two node NFS cross-mounting
Two node NFS cross-mounting

The two resource groups would be defined in SMIT as follows.

Resource group RG1 RG2
Participating node names Node A Node B Node B Node A
File systems The file systems to be locally mounted by the node currently owning the resource group. /fs1 /fs2
File systems to export

The file system to NFS export by the node currently owning the resource group. The file system is a subset of the file system listed previously.

 /fs1 /fs2
File systems to NFS mount

The file systems and directories to be NFS mounted by all nodes in the resource group. The first value is NFS mount point. The second value is the local mount point.

 /mnt1;/fs1 /mnt2;/fs2
File systems mounted before IP configured true true

In this scenario:

  • Node A locally mounts and exports /fs1, then over-mounts on /mnt1.
  • Node B NFS-mounts /fs1, on /mnt1 from Node A.

Setting up a resource group like this ensures the expected default node-to-node NFS behavior.

When Node A fails, Node B closes any open files in Node A: /fs1, unmounts it, mounts it locally, and re-exports it to waiting clients.

After takeover, Node B has:

  • /fs2 locally mounted
  • /fs2 NFS-exported
  • /fs1 locally mounted
  • /fs1 NFS-exported
  • service1:/fs1 NFS mounted over /mnt1
  • service2:/fs2 NFS mounted over /mnt2b

Both resource groups contain both nodes as possible owners of the resource groups.