Introduce nosync to speed up installability

[thrix]

This speeds up most of the packages significantly, I would consider including it:

zabbix-6.0.33-2.fc42                17 min      ->     13 min
kernel-6.10.6-100.fc39              1 hr 7 min  ->     1 hr 14 min
google-noto-fonts-20240701-1.fc41   4 hr 57 min ->     2 hr 27 min

[thrix]

@msrb PTAL pls

[rwmjones]

nosync is a no-brainer, provided you don't care about the results in the case of system failure. For CI that's almost always a reasonable assumption. In other words, it's a good idea.

[ppisar]

Disabling sync for CI environment makes sense because we are not interested in resuming a test after a CI crash.

[nirik]

Does nosync work again? https://bugzilla.redhat.com/show_bug.cgi?id=2019329

[thrix]

Does nosync work again? https://bugzilla.redhat.com/show_bug.cgi?id=2019329

I run the testing on all supported Fedoras to confirm it does not cause a failure of the pipeline:

https://osci-jenkins-1.ci.fedoraproject.org/job/fedora-ci/job/installability-pipeline/view/change-requests/job/PR-41/

[thrix]

Thanks all for the input. I will merge this once the above tests pass.

[thrix]

ok, more work needed in the plan, as my current testing showed nothing (rsync was not enabled)

[msrb]

I suspect that this env var is only set in the prepare test, and not in the subsequent installability test -- tmt likely spawns a new shell process for that.

[msrb]

I am wondering if it would be better to simply enable nosync in mini-tps directly?

[xsuchy]

We have been using nosync in Mock for ages. It causes mostly problem when you mix architectures. In this scenario this be no brainer. I can only imagine theoretical issues with some qemu tests or multiarch packages (is there any remaining? I guess not).