Feature request: Reboot required

[upnort]

Certain Proxmox packages do not notify administrators of a required reboot. For example, the kernel, glibc, libc6, systemd. Likely there are other packages but those come to mind.

With Debian and Ubuntu, such packages create /var/run/reboot-required.pkgs or /var/run/reboot-required. Those files provide administrators a way of evaluating reboots.

As /run typically is mounted in tmpfs, the files disappear on a reboot.

This is a handy feature. I hope you will consider. I can help test.

Thanks much.

 

[devinacosta]

This is not the right channel for feature requests. Please go here:
Please write your Request here: https://bugzilla.proxmox.com/

 

[fabian]

Certain Proxmox packages do not notify administrators of a required reboot. For example, the kernel, glibc, libc6, systemd. Likely there are other packages but those come to mind.

With Debian and Ubuntu, such packages create /var/run/reboot-required.pkgs or /var/run/reboot-required. Those files provide administrators a way of evaluating reboots.

As /run typically is mounted in tmpfs, the files disappear on a reboot.

This is a handy feature. I hope you will consider. I can help test.

Thanks much.

except for the kernel, none of those are provided by Proxmox. there are already lots of tools that handle stuff like this (I would recommend needrestart, but you probably want to add lxc-monitord and pve-container@* to the whitelist of services which should not be restarted).

 

[upnort]

Thank you for the reply.

If the kernel packages are updated to provide the standard Debian/Ubuntu /var/run/reboot-required.pkgs or /var/run/reboot-required, then that would be nice too.

I use the needs-restarting and needs-restart tools in CentOS. I was unaware of a similar tool in Debian. Thank you for that information.

 

[Sebastian Schubert]

Hi there,
any news on that - shouldnt be a big deal to provide those files with your custom rolled kernel.
We could really use these information for our ansible scripts to automagically reboot the systems, as we try to manage all our PVE nodes with ansible.

 

[Stoiko Ivanov]

any news on that - shouldnt be a big deal to provide those files with your custom rolled kernel.

Thanks for the ping - I created a bugzilla entry for the topic :

https://bugzilla.proxmox.com/show_bug.cgi?id=3528

subscribe there to get notified when we implement it (no promise when this will happen though)

 

[its-pve]

Now 3 years have passed, but nothing seems to have happened. We also use a monitoring tool for our various customers, which checks for necessary reboots and cannot currently detect the necessary reboot after a PVE kernel update.

Is there still a change, as something is happening here? Otherwise I would look for a workaround...

 

[LnxBil]

For years, we check if the next boot kernel is the same as we have booted right know and if they differ, there is a new kernel present and should be booted. This way, we always get notified and this is independend of any reboot-required files.

grep vmlinuz /boot/grub/grub.cfg  | head -1 | awk '{ print $2 }' | sed -e 's%/boot/vmlinuz-%%;s%/ROOT/pve-1@%%'

check that against

uname -r

 

[IIEP_IT]

I was also looking at this and I think a clean way I found of doing it is using ansible.builtin.apt in check_mode first:

    - name: Running apt in check mode to see what will be updated
      ansible.builtin.apt:
        update_cache: true
        upgrade: dist
      check_mode: true
      register: rc_apt

    - name: Flag a kernel upgrade
      ansible.builtin.set_fact:
        pve_kernel_upgrade: true
      when: '"Inst proxmox-kernel" in rc_apt.stdout'

Then farther down your play, just check for pve_kernel_upgrade to decide whether you need to reboot.