EOL notification system required for HWE kernels in LTSs

Landscape ought to detect that a system has an EOL HWE kernel installed and suggest appropriate remediation.

Consider a debconf high priority notice on the final update (postinst?). Or something else that will generate a root email, perhaps?

A root email isn't enough for desktop users.

I believe at the last sprint we had discussed having update manager inform the user, and to have the kernel available as an update in update-manager but unchecked by default, so the user needs to opt-in to it.

Canonical's OEM preinstall team would like to have a way to control the timing of the update. One way to do this is to provide a way to opt-out of the default upgrade notification, such as by dropping a file on disk via a software update delivered only to those OEM preinstall machines (we have a way to do this already). The second part of the request is a way to opt in/trigger the notification to the end user.

The reason is that Canonical in conjunction with OEMs will do upgrade tests on preloaded hardware to ensure that end users with this hardware have a smooth upgrade path to the next LTS kernel. We'd like to delay the upgrade notifications until those upgrades are tested.

It might also be worth considering what users already use to monitor their machines for updates, and to provide alerts by patching those mechanisms. For example, we might SRU a patch to the check_apt nagios plugin to pick up the EOL kernel state and present it as a warning or failure.

What other mechanisms are in use today?

Status changed to 'Confirmed' because the bug affects multiple users.

Am 30.01.14 20:31, schrieb Robie Basak:
> It might also be worth considering what users already use to monitor
> their machines for updates, and to provide alerts by patching those
> mechanisms. For example, we might SRU a patch to the check_apt nagios
> plugin to pick up the EOL kernel state and present it as a warning or
> failure.

Depending on such an implementation (for example adding those feature
with a switch) we may include this patch in Debian.

So, if all these update notification things check a single semaphore file, say "/var/lib/update-notifier/no-hwe-oel-warning", then the OEM people can drop that flag in place on a forced update, to avoid their users being notified of stack upgrades before they've validated the upgrade path.

Should we include the name of the base OS in the semaphore, e.g. /var/lib/update-notifier/no-12.04-hwe-eol-warning ?