Binary package hint: cloud-init
I'm investigating backporting support for guest managed kernels into 10.04 guest UEC images.
Doing so would allow 10.04 guest images that were running on EC2 and 10.10 UEC hosts to run 'apt-get dist-upgrade && reboot' and boot into their new kernel. Currently, the 10.04 guests cannot manage their own kernel under either UEC or EC2.
The solution in place for 10.10 is to use grub-pc to load the kernel on UEC and legacy-grub-ec2 to load the kernel on EC2.
In order to do this, we will need to pull the legacy-grub-ec2 package (a binary package built from cloud-init's source) to lucid.
I've done that backport and have a ppa build available at https://launchpad.net/~smoser/+archive/lucid-kernel-upgrades .
ProblemType: Bug
DistroRelease: Ubuntu 10.04
Package: cloud-init 0.5.10-0ubuntu1.3
ProcVersionSignature: User Name 2.6.32-309.18-ec2 2.6.32.21+drm33.7
Uname: Linux 2.6.32-309-ec2 i686
Architecture: i386
Date: Thu Nov 4 18:39:55 2010
Ec2AMI: ami-480df921
Ec2AMIManifest: (unknown)
Ec2AvailabilityZone: us-east-1d
Ec2InstanceType: t1.micro
Ec2Kernel: aki-6603f70f
Ec2Ramdisk: unavailable
PackageArchitecture: all
ProcEnviron:
PATH=(custom, user)
LANG=en_US.UTF-8
SHELL=/bin/bash
SourcePackage: cloud-init
===== SRU Information =====
* impact of the bug: This is not so much a bug as a missing feature from 10.04. With current 10.04 based EC2 images, an instance was not able to perform a kernel upgrade with 'apt-get dist-upgrade && reboot'. This is because the linux kernel is loaded by the hypervisor rather than a bootloader whos configuration is controlled by the instance. Near the end of the 10.04 cycle, Amazon announced "pv-grub" support. This allowed instances to manage their own kernel by managing /boot/grub/menu.lst
* how the bug has been addressed: In maverick, the grub-legacy-ec2 package was added under the cloud-init source package. This package manages /boot/grub/menu.lst much the same as grub-0.97 would (it is primariliy a source copy).
* minimal patch: The changes are available in my branch at [1], the diff can be seen from tip of that branch to revision 16 (current lucid-updates) [2].
* instructions on how to reproduce the bug:
Two bugs that are fixed by suggested changes:
1. system will not boot given a pv-grub kernel
* launch a lucid instance with pv-grub kernel
ec2-run-instances --instance-type t1.micro --kernel aki-407d9529 ami-b2b443db
2. system will not reboot into new kernel
* launch a lucid instance
ec2-run-instances --instance-type t1.micro ami-42887c2b
* install a newer kernel
sudo apt-get update && sudo apt-get dist-upgrade && sudo reboot
* reconnect, and verify old kernel is used.
* regression potential: The possibility for regression is centered around the fact that this change will cause
* new ubuntu images to load with a ramdisk
* kernel to boot with console=hvc0 parameter (used currently in maverick)
There is *no* potential for regression on running systems, only on systems where a user chooses to install grub-legacy-ec2 *and* modifies their instance to use pv-grub.
--
[1] https://code.launchpad.net/~smoser/ubuntu/lucid/cloud-init/lucid-kernel-upgrades
[2] http://bazaar.launchpad.net/~smoser/ubuntu/lucid/cloud-init/lucid-kernel-upgrades/revision/18?remember=16&compare_revid=16
=====
Marking fix-released. maverick has this already.