vmdb2 seems to usually work now, and that's been the situation for a while. Is it production ready? In my younger days, I would say yes, but now I'm not so sure. Reviewing the current roadmap, and thinking about it, I think the following are needed to declare it production ready (meaning the 1.0 release):

  • it's actually used in production
  • the test suite is of sufficient quality that I'm confident the software releaseable when the tests pass
  • the full test suite is run by CI; this includes actually building images and checking they run
  • the documentation is good enough that people can use the software

More concretely, I need to achieve these:

  • CI builds and tests images on each vmdb2.git change.
    • smoke.sh right now, but this needs to be extended to toy images that use all the plugins. Possibly on more than one architecture.
  • CI builds, tests (like smoke.sh does), and publishes arm (Raspberry Pi) and x86 images regularly.
    • Not just on vmdb2.git changes, as tools that vmdb2 uses may have changed, especially on sid.
    • I don't expect anyone to use these in anger, but the process exercises and validates vmdb2
    • For the arm images, Gunnar tells me they work for him. I expect him to do the official builds himself, but the CI should be useful anyway.
  • All the CI jobs run on both Debian buster and sid.
  • I have a version of v-i that can install Debian on a Thinkpad x220 laptop in a way that results in something I actually use in anger.

I don't have a good idea for how to validate that the documentation is up to snuff, so I'll punt on that and just assume it is, and improve it if and when people complain.