-
Notifications
You must be signed in to change notification settings - Fork 687
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Upgrade boxes for 1.8.0 (Xenial) #5870
Conversation
d23d77e
to
56a3883
Compare
rebased on lastest |
56a3883
to
a507bc1
Compare
On Debian-10, |
Steps required to get these boxes working on Debian 10:
|
Ran through upgrade scenario once more - this time submitting via the SI before running
|
Had some trouble running through the test plan myself. It looks like s3 is returning an HTTP 416 on the mon box, although if I use curl it fetches fine, with the sha256 checksum added here. At this point I think I'd prefer to rebuild and reupload the boxes and see if that helps, but I can't point to a more specific failure. Haven't investigated the GUI driver problems that @zenmonkeykstop reported, I'm not sure those are problems on my system currently. Will return for a closer look here. |
That 416 is a range not supported error from the server hosting the boxes. Nuke the contents of ~.vagrant/tmp/ and you should be good to restart. I think the gui stuff was a red herring and the boxes were stalling for some other reason. On latest Vagrant they complete the boot and are accessible via SSH, tho the gui is still frozen. |
Pushed up a commit cleaning up the 'upgrade' scenario a bit, addressing a few of the frustrations I had debugging. Mostly it's an eye toward #5512, which will largely be a copy/paste job from the Xenial upgrade logic. These changes don't solve the core problem with this PR, which is that the upgrade boxes as advertised don't work reliably. I believe the problem is that I intentionally fell back to an older version of the box #5794 (comment). In retrospect, I should have modified |
Sadly, doesn't resolve. I'm still seeing the box hang on first boot, during initialization of the display driver, same as @zenmonkeykstop originally reported. I've pushed up a commit with new checksums anyway, since the prepare logic is going to be required in the future, or by other builders. Make sure you remove any 1.8.0 boxes you may have locally to use the latest version. At this point, I suggest we step back from 1.8.0 and confirm that 1.7.1 upgrade boxes work well from the develop branch, to determine where the breakage lies. @rmol since I'm out tomorrow, perhaps you can lend a hand debugging here? |
@conorsch Yeah, of course. |
I didn't get through either the plain old process on I'll pick it up again tomorrow. |
I tested on develop just now, falling back to 1.7.1, and was able to create the boxes just fine. I'll open an issue to track, since @rmol reports different results. |
Removes the manually prepared "create" and "destroy" playbooks, since Molecule will handle those steps automatically based on the platforms declared in "molecule.yml". Try setting up a fresh Molecule scenario on v3 Molecule to observe the latest scenario skeleton: molecule init role --driver-name vagrant foo Updates the custom box logic to make sure the boxes are present locally, since Vagrant was blithely provisioning 1.7.1 upgrade boxes when I expected 1.8.0 locally.
We've had some trouble running the 1.8.0 upgrade boxes. This rebuilds them, using the latest available bento/16.04 version (202103.19.0), as well as a small tweak to the prep logic to provide a depedency that was recently removed from the Bento sources.
Debugging problems with upgrade scenario, where the app-staging VM hangs during cirrus video driver config stage. The staging scenarios set virtio video mode explicitly, so trying the same here. Has not resolved the problem, but might as well keep it for consistency's sake.
1d10b71
to
536176e
Compare
Was not able to get the new boxes working on my end. Tacked on a commit forcing virtio video drivers, which lets the VM boot past the cirrus stage where it was getting stuck before, but vagrant still marks the machine as in "pending" state indefinitely. I'm able to open a console to it, log in, and make network calls. At this point, I suggest we stop debugging the Xenial 1.8.0 boxes and work on writing a fresh upgrade scenario that supports Focal. |
Closing, but keeping the branch around since some of the work will be relevant for Focal. Will report progress over in #5512. |
Status
Ready for review
Description of Changes
Closes #5794
Creates upgrade boxes for 1.8.0, Xenial-only. Refs #5794. We should create Focal boxes too, but that's tracked in #5512 .
Testing
make build-debs
make upgrade-start
(will take a while as it needs to fetch the boxes)make upgrade-test-local
completes without errorDeployment
None, dev-only.