[mythtv-users] Diskless frontend - NFS options?

Brandon Penglase da5id2001 at SpaceServices.net
Sun Jan 17 02:27:09 UTC 2010


I have two frontends that netboot via gPXE, and actually load the
kernel over http, then filesystem on NFS. Here is the file that resides
on the webserver to tell the frontends what to load (think the pxelinux
config file):

#!gpxe
kernel http://10.0.3.2/bzImage-2631Z9-AXP-1 ip=dhcp root=/dev/nfs \
nfsroot=10.0.3.2:/nfsroot/fe2,nfsvers=3,tcp,hard,intr,nolock vga=0x315 \
splash=silent,theme:MythTV quiet console=tty1 
initrd http://10.0.3.2/fe1-initrd.img 
boot

Kernel to initrd is one line, no line breaks (email client is wrapping
it). In this case I also have a bootsplash setup (MythTV Splash!),
otherwise you can exclude the vga, spash, console and initrd.

Then I also have the same options in /etc/fstab (or else I got errors
and it didn't mount right.

fstab:
10.0.3.2:/nsfroot/fe1   /               nfs
rw,hard,nolock,intr             0 0

And note: 10.0.3.2 is my server, which does mythtv backend, dhcp, http,
etc, all the stuff. 

	Hope This Helps,
	   Brandon Penglase


On Fri, 15 Jan 2010 15:31:46 -0800
Jim Stichnoth <stichnot at gmail.com> wrote:

> On Mon, Jan 11, 2010 at 1:42 PM, Gareth Glaccum
> <gareth.glaccum at btopenworld.com> wrote:
> > Add a -hard to it? This would hang the system, whilst it
> > continuously tries to reconnect.
> > I have moved away from NFS on my frontends (I also run a full Gnome
> > session, lots of small files, so you may not have this problem), I
> > am using iscsi which seems a lot faster, not amazingly resilient,
> > it can handle a quick reboot of the server.
> 
> /proc/mounts shows that it is already using the "hard" option.  The
> notable difference between the NFS root file system and any other NFS
> file system is that the root file system is using the udp protocol
> whereas the others are using tcp.  I don't know if udp has anything to
> do with the problem, but I can't check since I haven't found any way
> to get the root mounted with tcp.  Any idea whether that could be
> related to the problem?
> 
> (To recap, it used to be the case that the NFS server could go down
> for a long time and the 3 diskless ION frontends would do no worse
> than hang until the NFS server was back up, after which the frontends
> would be just fine.  But something has changed and now the frontends
> have to be hard-rebooted if the NFS server goes down.)
> 
> Jim


More information about the mythtv-users mailing list