Boot image vs PXE install

[Robert Moskowitz]

The thread from Tomasz is interesting.  I was jsut in the mode of 
thinking that armv7 builds would be copy an image to SD and go.

I mean how much time to you save on an PXE build to SD versus xzcat to 
SD?  I might think that for the basic build, xzcat on a host is faster.  
Then you run a script that does what kickstart does.  The only 
consideration I have so far with this is sizing the partitions on the SD 
card.  I would like a larger swap to prepare for when I can hibernate to 
swap and getting the / partition to fill the card.

But then I started working with the partitions on the HDD.  It is a 
little harder to move a HDD between systems than a mSD card, so minimum 
boot from an SD card that then did the kickstart to the HDD (or SSD) 
makes more sense.

I can pull the F22-arm and repo to my local server; I do that for a 
number of distros every night (rsync changes) and do testing of a local 
install that way.  Of course my repo is on a different subnet than my 
install testing; I would have to put up the PXE boot function, etc.

So, lending a hand on a more Anaconda-style install to HD from a minimal 
SD card is something that I can help with, point me to what I can do to 
help.

[Paul Whalen]

----- Original Message -----
>  The thread from Tomasz is interesting.  I was jsut in the mode of
>  thinking that armv7 builds would be copy an image to SD and go.
>  
>  I mean how much time to you save on an PXE build to SD versus xzcat to
>  SD?  I might think that for the basic build, xzcat on a host is faster. 
Speeds can vary from board to board. Customization is the largest benefit.

>  Then you run a script that does what kickstart does.  The only
>  consideration I have so far with this is sizing the partitions on the SD
>  card.  I would like a larger swap to prepare for when I can hibernate to
>  swap and getting the / partition to fill the card.
>  
>  But then I started working with the partitions on the HDD.  It is a
>  little harder to move a HDD between systems than a mSD card, so minimum
>  boot from an SD card that then did the kickstart to the HDD (or SSD)
>  makes more sense.
>  
>  I can pull the F22-arm and repo to my local server; I do that for a
>  number of distros every night (rsync changes) and do testing of a local
>  install that way.  Of course my repo is on a different subnet than my
>  install testing; I would have to put up the PXE boot function, etc.
>  
>  So, lending a hand on a more Anaconda-style install to HD from a minimal
>  SD card is something that I can help with, point me to what I can do to
>  help. 
Installation testing is greatly appreciated. You will need a tftp boot 
server setup, rsyncing the entire repo isn't required but will speed
up the installations. 

The tests can be found here[1]. Installation on the development boards
does require a couple of manual steps at the end of an interaction 
installation or a section in %post to deal with the 'fdtdir' and re-write
uboot if you formatted the mSD/SD during installation.

Paul 

[1] - https://fedoraproject.org/wiki/Test_Results:Current_Summary

[Robert Moskowitz]

OK.  I will give this a look later today.  Got some people over the pond 
waiting for a report from me first...

On 03/09/2015 09:53 AM, Paul Whalen wrote:
>  ----- Original Message -----
> > The thread from Tomasz is interesting.  I was jsut in the mode of
> > thinking that armv7 builds would be copy an image to SD and go.
> >
> > I mean how much time to you save on an PXE build to SD versus xzcat to
> > SD?  I might think that for the basic build, xzcat on a host is faster.
>  Speeds can vary from board to board. Customization is the largest benefit.
>
> > Then you run a script that does what kickstart does.  The only
> > consideration I have so far with this is sizing the partitions on the SD
> > card.  I would like a larger swap to prepare for when I can hibernate to
> > swap and getting the / partition to fill the card.
> >
> > But then I started working with the partitions on the HDD.  It is a
> > little harder to move a HDD between systems than a mSD card, so minimum
> > boot from an SD card that then did the kickstart to the HDD (or SSD)
> > makes more sense.
> >
> > I can pull the F22-arm and repo to my local server; I do that for a
> > number of distros every night (rsync changes) and do testing of a local
> > install that way.  Of course my repo is on a different subnet than my
> > install testing; I would have to put up the PXE boot function, etc.
> >
> > So, lending a hand on a more Anaconda-style install to HD from a minimal
> > SD card is something that I can help with, point me to what I can do to
> > help.
>  Installation testing is greatly appreciated. You will need a tftp boot
>  server setup, rsyncing the entire repo isn't required but will speed
>  up the installations.
>
>  The tests can be found here[1]. Installation on the development boards
>  does require a couple of manual steps at the end of an interaction
>  installation or a section in %post to deal with the 'fdtdir' and re-write
>  uboot if you formatted the mSD/SD during installation.
>
>  Paul
>
>  [1] - https://fedoraproject.org/wiki/Test_Results:Current_Summary
>

[Tomasz Urbański]

Robert,
Thank you for encouraging words. It really helps in the struggle. ;-)

Of course, I fully agree that it is not wise to construct PXE build when
you have one box. However, when you have almost ten of them in the home
lab, and you multiply it by OS annual releases plus periodical maintenance
refreshment cycles it becomes more reasonable to automate the process.
Furthermore there is a question of quality that I'm strongly interested in.
I mean that, when I automate configuration management details (at
installation phase <<with PXE>> and further at operations phase <<with
Puttet/SaltStack>>) I can be sure that there are no accidental differences
and misconfigurations. That's my idea-fix.
(I try to execute this policy in my home lab [mainly Atom and ARM
architecture based] as well as in professional multicore multiblade
environment.)

As far as I'm concerned, the best thing you could do is to publish a series
of blog posts describing your findings and procedure (...even for newbies).

I'm interested in every detail of your way of doing.

Regards
Tomasz


2015-03-09 14:25 GMT+01:00 Robert Moskowitz <rgm(a)htt-consult.com&gt;:

>  The thread from Tomasz is interesting.  I was jsut in the mode of
> thinking
>  that armv7 builds would be copy an image to SD and go.
>
>  I mean how much time to you save on an PXE build to SD versus xzcat to
>  SD?  I might think that for the basic build, xzcat on a host is faster.
>  Then you run a script that does what kickstart does.  The only
>  consideration I have so far with this is sizing the partitions on the SD
>  card.  I would like a larger swap to prepare for when I can hibernate to
>  swap and getting the / partition to fill the card.
>
>  But then I started working with the partitions on the HDD.  It is a little
>  harder to move a HDD between systems than a mSD card, so minimum boot from
>  an SD card that then did the kickstart to the HDD (or SSD) makes more sense.
>
>  I can pull the F22-arm and repo to my local server; I do that for a number
>  of distros every night (rsync changes) and do testing of a local install
>  that way.  Of course my repo is on a different subnet than my install
>  testing; I would have to put up the PXE boot function, etc.
>
>  So, lending a hand on a more Anaconda-style install to HD from a minimal
>  SD card is something that I can help with, point me to what I can do to
>  help.
>
>
>  _______________________________________________
>  arm mailing list
>  arm(a)lists.fedoraproject.org
>  https://admin.fedoraproject.org/mailman/listinfo/arm