[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: jenkins.d.n for kfreebsd, PXE boot, text-mode d-i



Hi Steven,

On Montag, 1. September 2014, Steven Chamberlain wrote:
> I've been looking for a way to automate kfreebsd d-i testing, but found
> a solution that will also give Jenkins some test coverage of text-mode
> d-i (whereas it only tests GUI installer at the moment I think?) and the
> PXE boot images (instead of just CD ISOs).  So it would be really nice
> if we could enable this.  Please let me know if this is possible.

absolutly! I'm excited to see you working on getting bsd tests working on 
jenkins.d.n - and improving the d-i tests too!
 
> I've attached, because I think it was clearer this way, diffs from
> existing files to the new shellscript and preseed file I've created, that:

actually, git branches would have been even better :-) Next time, please! ;)

git branches and splitted commits will also allow me to take some bits of your 
work while asking you to polish some others. So you and me can both see 
progress too instead of endless discussions only :)

regarding d-i-installation-pxe.sh per se: I agree d-i-installation.sh is a 
hack, but what you did is making the hack worse, by copying parts of it into a 
very similar script, instead of a.) enhancing the existing script to do what 
you need too or b.) splitting that script into two/three scripts by using 
functions properly.

> * currently supports only GRUB2 (which is used for kfreebsd and hurd
> only - the other PXE boot images are syslinux; should be able to add
> support for that later)

thats fine (given the above) - I'm just not sure I think its sensible to have 
PXE+CD jobs _now_. Once they are working I think its a good idea to add more 
and more tests, but as long as thats not the case it just eats ressources.

> * currently supports only kfreebsd (the hurd PXE boot image for sid
> doesn't boot at the moment anyway)

fine as long as we dont create said jobs.

> * given a netboot.tar.gz instead of an ISO file, extracts it in the
> working directory
> 
> * edits the grub.cfg - which is not that much of a hack, because it's
> what a site sysadmin must do to define preseed parameters anyway
> 
> * sends kernel messages and d-i syslog to the virtual serial port log
> 
> * only needs a guest VM with 256 MiB memory to run, and 2 GiB disk

thats rather a low mem and low disk scenario...

> * doesn't need to run in sudo - I'm wondering why the script I based
> this on, was doing that?

the whole script isnt run with sudo, just single commands are. 

> * should (when we've fix the d-i bugs) test-boot the installed system,
> but that's untested yet, and I also haven't implemented save_logs yet
> (it would maybe need root privileges on the host to be able to mount the
> UFS partition, e.g. `mount -o ro,loop,ufstype=44bsd,offset=$n`)

using sudo to mount this is no problem

> I could use some help to write the Jenkins jobfile;  I don't have an
> installation of Jenkins to try this on but have dry-run this script and
> preseed file in isolation, they seem to work already.

I think it be best if you could modify the existing debian kfreebsd jobs, 
instead of creating new ones.

> Thanks!

likewise :-)

Feel free to come by on irc and/or open a bug against qa.debian.org so we can 
properly track this!


cheers,
	Holger


Attachment: signature.asc
Description: This is a digitally signed message part.


Reply to: