Solaris NFS
Developed by Sun Microsystems in 1983 the main purpose of the Network File System (NFS) is to share file systems under hydrogenous environment of machine running with different platforms.
NFS versions
NFS uses the RPC protocol for communicating between the NFS server and NFS clients.
To determine what version and transport of NFS is currently available, run rpcinfo on the NFS server.
# rpcinfo -p | grep 100003 100003 2 udp 2049 nfs 100003 3 udp 2049 nfs 100003 2 tcp 2049 nfs 100003 3 tcp 2049 nfs 100003 4 tcp 2049 nfs
From the about output the second column shows the NFS version and the third column is the transport protocol.
Sun has implemented the following versions of NFS on it?s operating systems, for both client and server:
| Version | Released | Solaris release | Comments |
|---|---|---|---|
| NFSv1 | 1984 | Sun used NFSv1 only for in-house experimental purposes. It was never released outside of Sun. | |
| NFSv2 | 1989 (RFC1094) | SunOS 4.x, Solaris 2.0 onwards | In NFSv2 clients needed to wait ~30 seconds to access modified data from the server. Data transfer speeds between server and client ~8kb. Unable to share file systems size greater than 2GB |
| NFSv3 | 1995 (RFC1813) | default in Solaris 8, 9 & 10. | In NFSv3 the wait time is less than 30 seconds, data transfer increased to ~32kb and provided an unlimited file system share size. |
| NFSv4 | 2000 (RFC3010), 2003 (RFC3530) | introduced in Solaris 10 | NFSv4 provides state full connections, improved firewall support and has a reduced number of server side daemons allowing for easier troubleshooting. |
Files
| File | Description |
|---|---|
| /etc/dfs/dfstab | contains all shared file system at boot time (we manually edit the file) |
| /etc/dfs/sharetab | contains all exported file system to the kernel |
| /etc/rmtab | contains all remote clients who are accessing the NFS share |
| /etc/default/fs | contains default file system information |
| /etc/fstypes | contains all file system which are supported by the Solaris OS |
| /etc/nfs/nfslogd.conf | contains all nfs log messages |
| /etc/default/nfslogd | complete configuration information about NFS log messages for the nfslogd daemon |
| /etc/default/nfs | contain the complete information about nfs protocol and nfs daemons |
Daemons
| Daemon | Description |
|---|---|
| rpcbind | The RPC portmapper |
| mountd | This is an RPC server that handles file system mount requests from remote systems. It checks /etc/dfs/sharetab to determine which file systems are available for remote mounting and which systems are allowed to do the remote mounting. |
| nfsd | The nfsd daemon handles other client file system requests. |
| statd | This statd daemon works with the lockd daemon to provide crash recovery functions for the lock manger. It keeps track of the clients that hold locks on a NFS server |
| lockd | This lockd daemon supports record locking operation on NFS files. It will send locking requests from the client to the NFS server. On the NFS server, it will start local locking. |
| nfsmapid | Its responsible for mapping nfs uid and nfs gid |
| nfs4cbd | It?s a daemon for NFS version 4 which is call backs the request. If NFS server does not response for a specific period of time |
| nfslogd | This nfslogd daemon provides operational logging |
Note: nfs4cbd and nfsmapid — Solaris 10 only. Solaris 11 uses idmap daemon instead of nfsmapid.
Sharing from the NFS server
Solaris 8, 9 & Solaris 10
share -F nfs -o [options] <share-path>
For example: share /oradata mount point as r/w to hosts orasys1 and orasys2 only. We use:
# share -F nfs -o rw=orasys1:orasys2 /oradata
Solaris 10 with ZFS
Similar example but for Solaris 10 with ZFS file system:
# zfs set sharenfs="rw=orasys1:orasys2" orapool/oradata>
Solaris 11 with ZFS
Under Solaris 11, the ZFS syntax is different than Solaris 10 as we need to name the share whilst sharing it:
# zfs set sharenfs=on orapool/oradata # zfs set share=name=oradata,path=/oradata,prot=nfs,rw=orasys1,orasys2 orapool/oradata>
Mounting from the NFS client
To mount the NFS share manually on the NFS client, we use the syntax:
mount -F nfs -o [options] <nfs-server>:<mountpoint>
To make NFS mount points persistent across reboots, we add the entries into /etc/vfstab file on the NFS client. For example:
#device device mount FS fsck mount mount #to mount to fsck point type pass at boot options 192.168.1.12:/oradata - /oradata nfs - yes rw,bg,vers=3
Using dfstab
In using the /etc/dfs/dfstab file we make sure that shares are persistent across reboots. The basic syntax for an NFS entry is:
share -F nfs -o rw=<host> -d "<share-description>" <mount-point>
Once entries are added to the dfstab file, we use the shareall and unshareall commands to share/unshare the entries mentioned:
Start/Stop NFS server
Solaris 8 and 9
# /etc/init.d/nfs.server start # /etc/init.d/nfs.server stop
Solaris 10
# svcadm enable svc:/network/nfs/server:default # svcadm disable svc:/network/nfs/server:default
Troubleshooting NFS
Whilst there are many commands to gather NFS share information. Below are the 3 most common:
| Command | Description |
|---|---|
| share | Use the share command on NFS server to display the current shares with all the options |
| showmount -e | The showmount -e command contacts the mountd daemon and lists all the shares with options. |
| dfshares | The dfshare command displays the same information as showmount -e but in a different format. |
Common error while configuring NFS:-
- RPC binding failure
- RPC program not register
- Client reboots error
- Stale handles error
- Resource not found
- Mount error
Review my other post on Troubleshooting various NFS errors for more details.