Make automatic updates nicer

[morgaesis]

Thank you for contributing to rpm-ostree.

If you are adding functionality to tree composes, please add
a corresponding test to the compose-test suite. Similarly,
if adding a client-facing feature, consider the vmcheck
suite. Regressions fixes are also great candidates for new
tests.

If you're not sure where or how to add tests, don't hesitate
to ask for help from the maintainers.

Cheers!

[openshift-ci]

Hi @morgaesis. Thanks for your PR.

I'm waiting for a coreos member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

Details

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[cgwalters]

Unfortunately this won't do what you want here, because ultimately this just calls into rpm-ostreed.service.

I think by default one would need to use systemd drop-ins for rpm-ostreed.service locally to have the desired effect.

In theory perhaps though, we could better differentiate "background updates" from "foreground activity" and the daemon could automatically lower its priority.

Hmm. Possibly there's a way we could inherit niceness and other properties from this unit.

[morgaesis]

Wow, you're quick to reply 😮
Isn't the systemd service the right place to configure this? The service has a key made for this. Configuring "niceness" in the process itself is extra complexity, which would probably just end up calling nicectl from within anyway...

[morgaesis]

I moved the niceness key to the daemon. When would you want rpm-ostreed to run at 0 niceness? It's such a heavy command, that I would always like it to be nice. Running the command manually or when automatic updates are applied, my system stutters.

[cgwalters]

Does this really help you though?

I suspect the main problem is not CPU usage but I/O bandwidth for most people.

[morgaesis]

Indeed, for the most part. There are still CPU bound parts

Writing OSTree commit: rpm-ostree start-daemon is CPU intensive (1-2 cores, i guess)
When I tried adding a new package, rpm-ostree start-daemon was using ~180% CPU.

Then, Nice is maybe not the right solution. I just want my system to run smoothly when the updates randomly start.

[morgaesis]

Or what.... man pages suggest that setting nice for the process also makes the I/O nice
https://linux.die.net/man/1/ionice

[cgwalters]

I think what would make me most comfortable here is a design where the automatic update path only has Nice set. Implementing that is a bit tricky; by far the simplest would be adding a nice flag for all transactions, and having the automatic update one set it by default.

Basically I'm less certain that we want to default to doing Nice=10 for all user-interactive invocations.