xenial64 vagrant boxes are unavailable from Atlas

[Winslow Dalpe]

Vagrant boxes for ubuntu/xenial64 cannot be retrieved from Atlas for at least the following versions, but most likely others as well:

- 20160608.0.0
- 20160606.1.0
- 20160531.0.0

ubuntu/xenial32 vagrant boxes appear to be unaffected.

Steps to reproduce:

1. Make sure you don't have the ubuntu/xenial64 box already downloaded
   vagrant box remove ubuntu/xenial64 --all

2. Try to re-add the box
   vagrant box add ubuntu/xenial64

3. Optionally, try to re-add the box with a version
   vagrant box add ubuntu/xenial64 --box-version 20160608.0.0

Output:

==> box: Loading metadata for box 'ubuntu/xenial64'
    box: URL: https://atlas.hashicorp.com/ubuntu/xenial64
==> box: Adding box 'ubuntu/xenial64' (v20160608.0.0) for provider: virtualbox
    box: Downloading: https://atlas.hashicorp.com/ubuntu/boxes/xenial64/versions/20160608.0.0/providers/virtualbox.box
An error occurred while downloading the remote file. The error
message, if any, is reproduced below. Please fix this error and try
again.

The requested URL returned error: 404 Not Found

[Dan Watkins (oddbloke)]

Xenial Vagrant boxes are registered with Atlas to point at http://cloud-images.ubuntu.com/xenial/current/xenial-server-cloudimg-amd64-vagrant.box; a hiccup in our build system meant that that file did not successfully sync out from our build hosts.

I've kicked off another build, so this should be fixed by that, and we will be addressing that hiccup in the near future, so we shouldn't see repeats of this.

[Dan Watkins (oddbloke)]

xenial64 boxes are now back to being available. :)

[Winslow Dalpe (rwdalpe)]

Thank you! :D

[Winslow Dalpe (rwdalpe)]

Also, your comment has me wondering: do all Atlas xenial boxes point to the current one, regardless of version?

[Winslow Dalpe (rwdalpe)]

It would seem that once again the boxes are unavailable. Same reproduction steps.

[Winslow Dalpe (rwdalpe)]

Boxes are once again unavailable.

[Dan Watkins (oddbloke)]

OK, my remedial actions for this were undone by our automation (because I missed a step >.<), so I will have to re-address this. Apologies!

[Patrik Karisch (patkar)]

Boxes are once again unavailable.

[Dan Watkins (oddbloke)]

OK, they're available again. I'm going to mark this as Fix Committed; Fix Released seems like a little too much hubris.

[Dan Watkins (oddbloke)]

(The changes to our automation to avoid this happening again have landed, but we haven't been able to validate that they are working in production yet :)

[Winslow Dalpe (rwdalpe)]

So far so good at this time (07:30 PST). Thanks for working on this.

Just to confirm something from before: you meant that all Atlas xenial boxes wind up resolving to the current latest xenial box, correct?

[Dan Watkins (oddbloke)]

They do; we automatically clean up dailies regularly, and we haven't automated cleaning up Atlas images. To avoid them no longer resolving at some point, all of them point to /current/.

Please feel free to report a separate bug if you want this to change (but be aware that we probably won't get to it for a while :).

[Winslow Dalpe (rwdalpe)]

Thanks for the clarification! I will file that separately.

[Winslow Dalpe (rwdalpe)]

For anyone who comes across this in the future, it was filed at https://bugs.launchpad.net/cloud-images/+bug/1591340

[Dan Watkins (oddbloke)]

Right, the bug in our automation appears to be fixed, so this shouldn't happen again.

If anyone sees this issue again, please open a separate bug. :)