Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Already on GitHub? Sign in to your account

'terminated' and 'shutting-down' machines should not be bogus #165

Closed
mrflip opened this Issue Sep 23, 2012 · 7 comments

Comments

Projects
None yet
3 participants
Owner

mrflip commented Sep 23, 2012

'terminated' and 'shutting-down' machines should not be bogus -- it's a perfectly expected circumstance. As a developer it's annoying -- leaving them marked as bogus means folks will get in the habit of adding --force, which is dangerous.

The previous version simply ignored them, which has always been fine. I suggest we go back to that behavior.

Contributor

temujin9 commented Sep 24, 2012

Agreed: known (minor) issue, and if you can fix this without breaking other parts of the bogosity, please do.

Owner

mrflip commented Sep 24, 2012

The main thing I need your guidance on is whether:

  • a) 'terminated' and 'shutting-down' machines do not exist. They are rejected at the discovery phase and so nobody is ever aware of them.
  • b) 'terminated' and 'shutting-down' machines don't exist, but you can ask about them: they are ignored during discovery unless the action requests their inclusion -- so knife cluster show would tell you about them, nothing else would.
  • c) 'terminated' and 'shutting-down' machines exist, but various actions ignore them

I prefer (a) -- by the time a machine is in that state, there's nothing you can do about it, and in some number of minutes it will be gone anyway, and I don't believe in ghosts. It's also simple enough I can jump in and do it; (c) sounds like a lot of conditional logic in the wrong places. If you don't mind (a) I can push in a patch.

Contributor

nickmarden commented Sep 24, 2012

+1, and also I agree that (a) is the cleanest choice from a user's point of view. If people really care so much to look at their terminated or dying instances, there's always the AWS (or similar) interface.

Contributor

temujin9 commented Sep 24, 2012

I would strongly prefer (c), even if we can't resurrect the corpses, which is why I didn't just do (a) myself. It does seem to require ugly conditional logic, which is why I'd held off.

Contributor

temujin9 commented Sep 24, 2012

An acceptable compromise might be to add a flag (or interpret the existing -V flag) to knife cluster show to "display dead machines". That way, launch/etc. commands will ignore those machines, but I can still get my visibility without going all the way to the consoles.

Contributor

nickmarden commented Sep 24, 2012

I can be convinced of that as well, as long as I don't have to type --force (which I don't really ever want to type) in order to bring up a server that I recently terminated.

Owner

mrflip commented Sep 24, 2012

okey doke: knife cluster show will be the only one that brings up dead nodes, and only with verbosity >= 1.

@ghost ghost assigned mrflip Sep 24, 2012

@temujin9 temujin9 closed this Oct 1, 2012

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment