[GTALUG] simple straightforward tool for mainaining small networked of imaged linux boxes

Christopher Browne cbbrowne at gmail.com
Tue Oct 7 18:57:37 UTC 2014


On 7 October 2014 13:18, Lennart Sorensen <lsorense at csclub.uwaterloo.ca>
wrote:

> On Tue, Oct 07, 2014 at 07:01:42AM -0400, Matt Price wrote:
> > Lennart, can you explain that process in more detail?  I mean, what tool
> > would one PXE boot from?
> >
> > Walter, that is helpful but still sounds pretty time-consuming -- I would
> > like to find a way to have an unattended re-imaging that can run
> > more-or-less simultaneously on all the machines.
>
> If the BIOS has an option for network boot, then you could enable that.
> Then the server would have to have 'pxe' (pxe boot daemon) running to
> send the boot files to the client when it powers on.  This would then
> provide the boot menu and such for the client, which could normally be
> to load the kernel and ramdisk from the local disk, or to network boot
> a kernel and ramdisk that reflash the local disk, perhaps using partimage
> to transfer the image from the server to the client.
>
> You might even set up wake on lan if supported so you could remotely
> turn the machines on when you want them to do the reimaging of the disk.
> In fact a lot of machines can have a different boot order for normal
> boot versus wake on lan boot making some parts even simpler.


Scott Sullivan did a talk on setting up PXE boot a couple of years ago.  I
recently pointed back to the memories when I put in a new media
machine (Zotac ID-88), and found it wasn't recognizing bootable CDs,
only to realize that some of the messages in the failing boot indicated
a "looking for PXE" step.

I was keen on putting Debian in place, so the following instructions were
relevant to that:
   https://wiki.debian.org/PXEBootInstall

The somewhat oversimplifcation is that you need to set up several things
on a central server:
a) Modify DHCP to indicate the offering of PXE boot
b) Install tftp on the server, and put install files in a suitable place
    so clients may pull them

The "most magical" part, which is open-ended as to how sophisticated
it can get, is the part of controlling what gets installed.

In my "wanna install Debian" case, it was simple enough; I grabbed a
stock Debian netboot install and threw that into place.  I didn't much
care how time-consuming later steps were.  Presumably this means I'll
not likely be pulling many Debian CD images anymore; using PXE is
about as easy, and simpler as it consumes no CDs/DVDs.

If you want to have a bunch of MINT machines with extensive common
configuration, then it's worth jumping through some hoops to customize
the net install so it needs little work later.

There are some instructions on this for MINT:
   http://forums.linuxmint.com/viewtopic.php?f=197&t=96322

It's worth your while to look into some automation tools for installing
common things that need to be on each machine.  My answer is
cfengine2 (I haven't migrated to v3); others like Chef/Puppet/Ansible.
My sense is that it's likely preferable to have a fairly simple
installation, and use tools to do system management afterwards.

I know Scott has gone through the notion of having scripting designed
to customize Fedora/CentOS installations; that's a different
approach, but certainly valid.  If you wanted MINT, then his scripts
are likely useless to you, alas.
-- 
When confronted by a difficult problem, solve it by reducing it to the
question, "How would the Lone Ranger handle this?"
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://gtalug.org/pipermail/talk/attachments/20141007/aa5e1012/attachment.html>


More information about the talk mailing list