search

projectatomic / commissaire-mvp

A lightweight REST interface for upgrading, restarting, and bootstrapping new hosts into an existing Container Management cluster.
http://commissaire.readthedocs.org/en/latest/
GNU General Public License v3.0
15 stars 9 forks source link

[merged] Vagrant file and updates #163

Closed ashcrow closed 8 years ago

ashcrow commented 8 years ago

This change adds a Vagrantfile, a new option for bdd tests to use Vagrant IPs, and updates documentation. The Vagrant set up creates the following hosts:

Example:

   $ vagrant up
   ...
   behave -D use-vagrant
   ...
ashcrow commented 8 years ago

Tagging @cooktheryan just so he knows about this change when get gets back :smile:

ashcrow commented 8 years ago

:arrow_up:

mbarnes commented 8 years ago

Fixups look good. :+1:

I'm trying out the Vagrantfile. After letting it run for some time I came back and found all my VMs stuck in "paused" while Vagrant was trying to upgrade fedora-cloud and they wouldn't resume. The log showed it had just upgraded ssh, but dunno if that had anything to do with it. Destroyed the machines and trying again while watching it closer.

In any case, this is freaking awesome! Amazed you got this working already.

mbarnes commented 8 years ago

So I've seen this behavior a few times now:

If I bring up just the first three machines (etcd, fedora-cloud, fedora-atomic) they do their upgrade and seem fine. I even let them run awhile just to be sure.

Then as soon as I start up the commissaire machine, both commissaire and etcd machines immediately lock up. (virt-manager shows their status as "Paused (I/O error)" and I don't know how to get more detail.) A short time later fedora-cloud also locks up for some reason.

The Vagrant log doesn't show any errors. I think I've seen it lock up before it even gets to the inline script, but this latest time it got as far as "Updating the system".

fedora-atomic has so far been the lone survivor.

Does this ring any bells for you?

ashcrow commented 8 years ago

Unfortunately no. I've tested on libvirt and with VirtualBox with all of the boxes and haven't had any lock up issues. I have seen upgrade take longer than I'd normally expect while provisioning but never any pause or lock up. Could it be a libvirt configuration item or not enough resources issue?

mbarnes commented 8 years ago

Resolved: NOTABUG Turns out I just have too many libvirt images for my poor unendowed root partition.

Sorry for the noise. Behave tests all pass with -D use-vagrant. :+1: :+1:

mbarnes commented 8 years ago

@rh-atomic-bot r+

rh-atomic-bot commented 8 years ago

:pushpin: Commit b69e2a2 has been approved by mbarnes

rh-atomic-bot commented 8 years ago

:hourglass: Testing commit b69e2a2 with merge a682821...

rh-atomic-bot commented 8 years ago

:sunny: Test successful - travis Approved by: mbarnes Pushing a68282193c413057db7003f040f62a2e76c9d26e to master...