Dell Update Packages (DUP) and LiveCD/DVD

Cal Webster cwebster at ec.rr.com
Wed Oct 19 09:04:38 CDT 2011


On Wed, 2011-10-19 at 14:09 +0200, Sven Ulland wrote:
> On 10/18/2011 08:39 PM, Cal Webster wrote:
> > I think the simplest option to give customers is the LiveCD (actuall
> > LiveDVD) you introduced. However, for it to be useful to the widest
> > costomer base it should be split up into a "Legacy LiveCD" ISO (less
> > than 700 MB to fit on CD-R) and "Current LiveDVD" (handling updates
> > for the newest servers in the lineup and maybe as many older systems
> > as would fit). Most of the older Dell servers (PE2600/2800/600SC)
> > only have CDROM readers and the system BIOS will not permit booting
> > from external USB DVD, flash, or hard drives. These systems cannot
> > take advantage of firmware updates unless they're running CentOS 5,
> > which they are not. An improvised CentOS LiveCD such as we used is
> > the only viable option for these legacy systems.
> 
> Do they support pxe booting? Having a solid pxe infrastructure with
> flexible images/systems to boot is so much more useful than the livecd
> approach :) LiveCDs that run directly from their initrd can also be
> put directly on pxelinux, without modifications.
> 
> On that note, I don't know why Dell (or HP, etc) don't focus more on
> pxe-based deployment approaches. Imagine being able to download the
> latest firmware update image, schedule reboots of your systems
> (including setting pxe as a one-time boot option), booting them on the
> image (with pxelinux.cfg per mac or ip address), and having them
> upgrade themselves and reboot automatically. Progress or errors would
> be sent via syslog or snmp, and modifying the image would be a breeze
> as long as its built properly.
> 
> Doing this for an entire enclosure, rack or datacenter would be
> a breeze! :)
> 
> sven

I agree that PXE boot would be a valuable option, so long as the server
has a NIC that supports it. As you say, data center/rack deployments
might be ideal candidates. Many legacy systems are not equipped this
way. Also, the PXE infrastructure is not something everyone would want
to deploy to support it, whether for security reasons or resource
expenditure.

I would lean toward an RPM-based deployment. This would dovetail
seamlessly with existing update infrastructure using "yum" or "up2date".
For firmware/BIOS updates that can be deployed on an active system there
would be no down-time. For updates requiring a quiescent system, a
custom initrd could be installed as the default boot option to be
executed at the next reboot. When done, the initrd and other temp files
could be removed before an auto-reboot into the normal kernel. RPM / Yum
updates are much more effective in terms of configuration management,
dependency resolution, and software deployment than one-off tar-balls or
these un-managed DUP updates. RPM would keep a record of the updates in
the RPM database too, making it a simple matter to determine current
firmware versions.

./Cal



More information about the Linux-PowerEdge mailing list