NFS fails when using custom TCP/IP stack

Products

• VMware
• VMware vCenter Server
• VMware vSphere ESXi

Issue/Introduction

Symptoms:

• NFS traffic goes through the default stack even though default stack was created with a dedicated subnet.
• NFS traffic fails to use custom TCP/IP stack.
• Unable to add new NFS mounts with custom TCP/IP stack.

Environment

VMware vSphere ESXi 6.x
VMware vSphere ESXi 7.x

Resolution

To resolve this issue, firstly put the host into MM, and then specify the Custom Netstack in NFS services for NFS to use custom TCP/IP stack.

To specify the Custom Netstack in NFS services:

1. Run this command to create the netstack:

# localcli network ip netstack add --netstack=nfs

2. Run this command to unmount any existing NFS shares:

# localcli storage nfs remove -v volume-name

3. Run this command to stop the NFS service:

# /etc/init.d/nfsgssd stop

4. Run this command to unload the NFS modules:

# vmkload_mod -u nfs41client
# vmkload_mod -u nfsclient
# vmkload_mod -u sunrpc

5. Run this command to configure the sunrpc module with the specified netstack:

# localcli system module parameters set -m sunrpc -p netstackName=nfs

6. Run this command to reload the NFS modules:

# localcli system module load -m sunrpc
# localcli system module load -m nfsclient
# localcli system module load -m nfs41client

7. Run this command to restart the NFS service:

# /etc/init.d/nfsgssd start

8. Run this command to remount the NFS shares:

# localcli storage nfs add -v <Datastore_Name> -H <NFS_Server_address> -s </share_mount_point>

Example:

localcli storage nfs add -v NFS_Datastore -H xxx.xxx.xxx.xxx -s /NFS_share/esx_nfs01

Additional Information

Custom TCP/IP stack for NFS has been deprecated from vSphere 8.0 and the workaround to use Default TCP/IP stack to fix the issue.

Additional reference article:

Remounting a disconnected NFS datastore from the ESXi command line