Disable motd

[stijndehaes]

Disabling motd results in a faster boot of the
AMI.

Issue #, if available: #1751

Description of changes:

By disabling motd the AMI is able to boot up faster. The difference is about 20s. Analysis was done using systemd-analyze:

Before change:
Startup finished in 698ms (kernel) + 7.574s (initrd) + 1min 2.208s (userspace) = 1min 10.481s graphical.target reached after 1min 2.146s in userspace.

After change:
Startup finished in 454ms (kernel) + 3.842s (initrd) + 38.513s (userspace) = 42.810s graphical.target reached after 38.055s in userspace.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

Testing Done: Launching a custom AMI made from one of the release AMI's

See this guide for recommended testing for PRs. Some tests may not apply. Completing tests and providing additional validation steps are not required, but it is recommended and may reduce review time and time to merge.

[stewartsmith]

Is this observable with the stock AL2023 images? We did make sure the update-motd service didn't block the OS launching.

[cartermckinnon]

This service doesn't block any of the important services (nodeadm-run, containerd, kubelet), so I don't see much benefit in disabling it. The "update is available" message written by /etc/update-motd.d/70-available-updates is helpful, IMO.

Time to graphical.target isn't the right metric here, I think, we care about kubelet starting, the Node becoming Ready in the API, etc.

[cartermckinnon]

Let me know if you disagree @stijndehaes

[stijndehaes]

@cartermckinnon fair enough :)
I do notice kubelet launching 3s faster without motd, but that is probably within margin of error.
I would like to improve the hot code path, I am leaving on a 10 days of vacation tomorrow, but after that I would love to improve nodeadm-config and nodeadm-run.
nodeadm-config uses kubelet making it slow, and nodeadm-run uses containerd making it slow.
Is there a better place than github issues to discuss solution designs? :)

[ndbaker1]

appreciate the attention here, i also think it would benefit from a bit of caching and/or parallelism here so i might take a stab at it when i have time