My co-worker (and co-author) Ricardo Labiaga, just told me:
The NFS v4.1 client has been officially integrated into the upstream Linux kernel repository! It will be part of the Linux kernel version 2.6.31. We should see a release candidate before the end of this week. The Linux NFS v4.1 client implements the Sessions functionality, laying the foundation for pNFS to be submitted in the next few months.
This is a major milestone for the Linux NFS client development community (consisting of a group that includes developers like Ricardo from NetApp, EMC, IBM, Panasas, and the University of Michigan) and for NFSv4.1. Congratulations!
according to changelog NFSv4.1 session code is already a part of 2.6.30 :
http://kernelnewbies.org/Linux_2_6_30#head-41d9a4e3cf4ef893e8ff39b37811aa33373156fe
Hopefully, pNFS code will go into 2.6.31
Posted by: Tigran Mkrtchyan | June 23, 2009 at 11:02 PM
To expand on Tigran's comment... Only the NFSv4.1 _server_ session code went into 2.6.30. It is the client NFSv4.1 session support that has just been integrated into 2.6.31.
We will be working on the integration of the pNFS code in the next few months.
Posted by: Ricardo Labiaga | June 24, 2009 at 09:08 AM
Hi,
Out of curiosity, what kind of filesystem / disk caching setup does NetApp use internally? The reason I ask is because I've been considering running ESXi servers on top of NFS, but some sources (e.g. XFS FAQ) state that virtualization products virtualize disk writes in a way that prevent filesystem write barriers from working correctly. This of course means that either you need to turn off individual disk cache (and suffer the performance penalty), or risk filesystem corruption in the event of a power-out. Also, does VMFS somehow work around this problem?
Posted by: Jukka P | July 17, 2009 at 04:31 AM