User process freezes on NFSv4 client

User process freezes on NFSv4 client

[Gabor Kovacs]

Hello All,

I've got NFSv4 problems on a big partition. Both NFS server and client run 
kernel version 2.6.21.6. Exported file system is on an AoE etherdrive 
(/dev/etherd/e1.1) device, having no partition table, XFS file system. The 
file system is pretty large,  8.9T. Typical file size 1-10M. 

On the client side, after much time (3-7 days) of file intensive task, the 
user process freezes. Top command shows a "D" state for it, but many other 
command trying to "touch" it, freezes, too. (ps, fuser /nfs_dir, w) None of 
the frozen processes can be killed. New processes, however, can access the 
nfs4 mounted file system. Any idea?

Thanks
Gabor Kovacs

Re: User process freezes on NFSv4 client

[Andrew Morton]

> On Thu, 16 Aug 2007 18:19:14 -0400 Gabor Kovacs <kgabor@bolyai.elte.hu> wrote:
> Hello All,
> 
> I've got NFSv4 problems on a big partition. Both NFS server and client run 
> kernel version 2.6.21.6. Exported file system is on an AoE etherdrive 
> (/dev/etherd/e1.1) device, having no partition table, XFS file system. The 
> file system is pretty large,  8.9T. Typical file size 1-10M. 
> 
> On the client side, after much time (3-7 days) of file intensive task, the 
> user process freezes. Top command shows a "D" state for it, but many other 
> command trying to "touch" it, freezes, too. (ps, fuser /nfs_dir, w) None of 
> the frozen processes can be killed. New processes, however, can access the 
> nfs4 mounted file system. Any idea?
> 

Next time it happens, please run

	echo t > /proc/sysrq-trigger

then send the output of `dmesg -s 1000000'.  Thanks.