[mythtv-users] Diskless frontend - NFS options?

[Jim Stichnoth]

On Sun, Jan 17, 2010 at 2:12 PM, Kevin Ross <kevin at familyross.net> wrote:
> Jim Stichnoth wrote:
>>
>> Thanks, that helped me get to a PXE boot setup where the NFS root uses
>> TCP.  Unfortunately, it turns out to make no difference -- after
>> rebooting the NFS server, the frontend root file system is gone and
>> doesn't come back until the frontend is rebooted.  I wonder if the
>> same thing happens on your system when the NFS server goes down and
>> comes back?
>>
>> Jim
>
> I have two diskless frontends, and both recover just fine after rebooting
> the server. My config:
>
> LABEL linux
>       kernel vmlinuz-2.6.30-2-686
>       append rootfstype=nfs root=/dev/nfs initrd=initrd.img-2.6.30-2-686
> nfsroot=192.168.1.2:/raid/netboot/bedroom ip=dhcp rw
>
> I do NOT have the root filesystem listed in my /etc/fstab on the clients.
>  It doesn't seem to hurt anything not having it there.
>
> On my system at least (Debian Squeeze), it uses TCP by default, as well as
> hard mounting.  Looking at /proc/mounts:
>
> 192.168.1.2:/raid/netboot/bedroom / nfs
> rw,relatime,vers=3,rsize=262144,wsize=262144,namlen=255,hard,nointr,nolock,proto=tcp,timeo=7,retrans=10,sec=sys,addr=192.168.1.2
> 0 0

Thanks, it's good to hear that your NFS root recovers after a server
reboot.  My NFS options in /proc/mounts are pretty much the same
except that I have timeo=600,retrans=2 but I doubt that is at all
related.  I have all the necessary drivers built into the kernel so I
don't need an initrd image.

If I unplug the network cable and later reconnect it, the frontend
recovers fine.  The problem is with the NFS server rebooting.  After
it comes back up, the network activity LED on the frontend flashes
full-out, to the point where it looks almost solidly lit, so
apparently the frontend is trying really hard to remount something.
At the same time, nfsd processes on the server show a significant
amount of CPU time being consumed, which dies down to practically
nothing after the frontend is rebooted.  By temporarily enabling some
logging, I quickly get tens of thousands of log messages like this:

Jan 17 08:57:59 localhost kernel: RPC: Want update, refage=120, age=85

This must be a bug or incompatibility or something in my kernel, which
is 2.6.31.6, but it's surprising that I can't find any similar reports
in a web search.  I guess I could try a different kernel version.

Jim