We would need to point packages.flapjack.io to the URL above
Sources lists would need updating (perhaps this would fit in with the flapjack v2 release)
We would no longer have an experimental component, but a separate repository
We would no longer be able to sign the debian repository, so would need to sign the packages instead [2]
This has a 25 package limit, so we'd need to make sure we remove older point releases [3]. We'd archive all of these, once, in the s3 bucket (but not in a repository).
I don't believe that any of these are big issues if we combine this with the 2.0 switch. Is this something we want to try (particularly before we prune the aptly db again)?
[1] https://packagecloud.io/#features
[2] https://github.com/chef/omnibus/issues/402
[3] We support 2x Ubuntu, 1x Debian, 1x Centos, so we'd be able to have 6 releases per OS at one time. If we kept the last 2 main point releases, we could use 4 for testing. This seems reasonable, as long as we have the promote script remove packages relating to the package being promoted from testing.
Packagecloud.io looks very interesting [1], and would mean that we can deprecate our aptly and createrepo code and s3 files.
I've created a tentative repository at https://packagecloud.io/flapjack
This has the following issues:
I don't believe that any of these are big issues if we combine this with the 2.0 switch. Is this something we want to try (particularly before we prune the aptly db again)?
[1] https://packagecloud.io/#features [2] https://github.com/chef/omnibus/issues/402 [3] We support 2x Ubuntu, 1x Debian, 1x Centos, so we'd be able to have 6 releases per OS at one time. If we kept the last 2 main point releases, we could use 4 for testing. This seems reasonable, as long as we have the promote script remove packages relating to the package being promoted from testing.