[mythtv-users] Disk space for frontend-only machine

[Adrian Saul]

On 17/08/2012 4:09 PM, Phill Edwards wrote:
> I've never tried setting up a network boot. Isn't it slow to boot? If
> you have to transfer a few GBs of data for the MythTV frontend image
> doesn't that take a long time (even over a gigabit network it's going
> to take some time to transfer all that data)?
> _______________________________________________
> mythtv-users mailing list
> mythtv-users at mythtv.org
> http://www.mythtv.org/mailman/listinfo/mythtv-users

You don't have to copy the entire thing - the boot block and a miniroot 
via TFTP,  then once it mounts the NFS image it opens files as needed.   
You also get some benefits of your NFS server doing caching as well as 
your client machine.

I actually have my home setup running like this with a Solaris server at 
the back doing the storage (this lets me ZFS snapshot the image around 
upgrades etc).  My backend/frontend and a standalone frontend are both 
network boot.   Performance wise its fine, the NFS benchmark I have got 
is 100MB/s read and a little slower for write - I also have the database 
on the NFS server so I can snapshot it as well and so its isolated from 
the upgrade process.   The noticeable impact is latency around file 
attributes or copying lots of small files, that tends to be a lot 
slower.   Doing things like release upgrades tends to be a lot longer 
than off disk.  In my setup I also suspect some Solaris vs Linux NFS 
issues at play, but for the most part it just works and forcing things 
to NFSv3 seems to help.

Over the last two nights I have actually worked to upgrade my current 
setup from Mythbuntu 11.04 + 0.25.2 to 12.04 + 0.26.   What I did was 
use VirtualBox to to build the new backend and frontend images clean 
from ISO.  I did some tweaking to add extra packages and some known 
config, then copied it over to my NFS server using rsync.  I then set 
that up as a new boot image and booted the machines off the new image - 
bang - upgraded, with the old image there as a rollback if need be.    
Actually I did spend several hours after that sorting out various 
database, 0.26 and missing config issues, but thats all part of 
documenting the process for next time.