Re: [Drbl-user] insert alternative ethernet kernelmodules n drbl - clonezilla live-mode

Op vrijdag 01-07-2011 om 08:08 uur [tijdzone +0800], schreef Steven
Shiau:
> 
> On 06/27/2011 09:32 PM, koe...@te... wrote:
> > I'm setting up a drbl-server with clonezilla-live as 'back-end'.
> > 
> > I've had a problem (connection dropped after several minutes - half an
> > hour) with my ethernetmodules on my (laptop)server: the networkcard is a
> > realtek 8111 but the module used is r8169. To make the story short:
> > everything works fine when compiling the modules from the
> > realtek-website and blacklisting r8169.  You can find the guide I used
> > on http://bernaerts.dyndns.org/linux/49-debian-server-rtl8111
> > 
> >  
> > 
> > Now I thought everything should just work fine but... the same problem:
> > connection dropped after 10 minutes when trying to restore 4 laptops.
> > After a bit of analysing I found that the clients (also laptops) are
> > using the same networkcard, and yes they also use the wrong
> > ethernetkernel-module.
> > 
> > 1) How can I get the right kernelmodules in clonezilla-live (version 486)?
> Did you try to use Clonezilla live 20110623-natty as the back-end?
> Or try Clonezilla live 1.2.9-11.
> > 
> > 2) And -if this is possible- how can I get this remastered
> > clonezilla-live in drbl? 
> You can refer to this:
> http://drbl.sourceforge.net/faq/fine-print.php?path=./2_System/81_add_prog_in_filesystem-squashfs.faq#81_add_prog_in_filesystem-squashfs.faq
> 
> Steven.
> > 
> >  
> > 
> >  





I followed this suggestion. Unsquashed the iso and then chroot
in /tmp/clonezillan/... I eventually could compile the drivers from the
realtek-website inside the chroot-environment after installing several
extra kernel- and buildpackages. The real trouble was with Initramfs:
there was no /boot-directory. Which is true: the compiled kernelimages
should be outside of the chroot-environment. That's where I gave up (I
ran out of answers)...

I had the same issues even with the ubuntu version of clonezilla: it
also uses the r8169-kernelmodule instead of the rtl8111.
 
So how to solve this problem?  I had tested several types of laptops
with one-to-one connection when my server had no adjusted kerneldriver.
No switch or router involved so the connection was not interfered by a
non-working of defect switch. I remembered the time the connections were
up and the amount of datatransfer they had. On some laptops with other
ethernetcards I sometimes had speeds up to 2 GB per minute. The cloning
stopped after about 19 minutes. I  once had a connection of about 3GB
per minute: connection stopped after 15 minutes. And a real high speed
of 5 GB/minute stopped after about 3 minutes. There is a logic between
stopping of connection and dataspeed when using the wrong
kerneldriver.  
If I could limit the dataspeed, what would then happen? Would the
connection last longer? I looked around for the apropriate tool to do
this under linux: ethtool. After some tweaking I found this solution to
work:  
ethtool -s eth0 speed 100 duplex half autoneg off
In short: limit the speed to 100, no full duplex and autonegotiation
(networkcards are polling each other) is off. 
It works like a charm. Yesterday I cloned two laptops simultaneously
which would otherwise 'crash' all the time (CRC-error) . There is a
switch in between. Speed is about 750 MB/minute when you clone an almost
full partition. 

So my problem is solved. The solution is not clean but at least it
works. Maybe you could make it an extra topic in the FAQ: "possible
solution when connection drops" or " CRC-error message" . 
If there are problems, people tend to blame the software, while it might
also be the hardware or kerneldrivers. This also needs to be checked
when there are problems. 

  
Kind regards, 



Koen Wybo