Current status of multi-homes machines

Stephan Hermann sh at sourcecode.de
Mon Dec 6 15:45:36 CET 2010


Hi Carsten,

did you set IPAPPEND=2 in pxelinux.cfg/<pxefile> ?

without that BOOTIF doesn't work and won't be set by pxelinux

Regards,

\sh
On Thu, 2010-12-02 at 10:30 +0100, Carsten Aulbert wrote:
> Hi all
> 
> On Wednesday 01 December 2010 19:00:24 Carsten Aulbert wrote:
> 
> > > Use fai-chboot -P. The initrd code includes some code which evaluates
> > > the BOOTIF kernel boot parameters. fai-chboot -P in cooperation with
> > > pxelinux will set this variable.
> > 
> > This looks much more promising than Michael's proposal, especially if -
> > over time - a large number of machine which were more or less homogeneous
> > are becoming mire and more diverse.
> 
> * -P sets the correct variable - good
> * box still stalls
> 
> after 1 hour in the (noise) cluster room, some findings:
> 
> o the BOOTIF environment variable is set
> o setting 'ip' in the pxe config file does not seem to make any difference 
> anymore (in lenny it did)
> o weird part, after I changed the init script of the initrd in the NFSROOT to 
> output debug info "set -x" and outputting the environment variables "env", 
> update the existing initrd.img (still in NFSROOT) and copying it into the 
> tftp/fai tree, I still don't see this output while PXE booting the node
> o after entering the busybox after PXE boot, I do see my changes, I source the 
> functions file from scripts/ and run configure_networking by hand (and with 
> "set -x")
>   o the first part after 'if [ -n "${BOOTIF}" ]; then' works as expected, from 
> the BOOTIF variable the correct device is found
>   o since it's unconfigured, there is no /tmp/net-"${DEVICE}", thus it enters 
> the large foor loop ROUNDTTT
>   o IP is set to 4 values (<client-ip>:<server-ip>:<gw-ip>:<netmask>, not the 
> full 7 nor is is ip=dhcp regardless of what is in the config file on the tftp 
> server), therefore it starts ipconfig with -d $IP and stops there with eth0 as 
> there is no DHCP server
>   o I'm not 100% sure why the time out does not occur here, as the sum of all 
> these numbers is less than 300s. Thus after 1 hour I would expect anything to 
> happen here.
> 
> Open questions:
> 
> o Why no timeout?
> o Where is my output (console=tty0 is set)?
> 
> I'll try to cheat and just add my custom case here for now and see if that 
> works.
> 
> Cheers
> 
> Carsten
> 

-- 
Stephan '\sh' Hermann
SysAdmin / Ubuntu Developer
xmpp: sh at sourcecode.de






More information about the linux-fai mailing list