Open Stack

Excerpts from Gregory Haynes's message of 2014-10-01 19:09:38 -0700:
> Excerpts from Clint Byrum's message of 2014-10-02 01:50:33 +0000:
> > Recently we've been testing image based updates using TripleO, and we've
> > run into an interesting conundrum.
> > 
> > Currently, our image build scripts create a user per service for the
> > image. We don't, at this time, assert a UID, so it could get any UID in
> > the /etc/passwd database of the image.
> > 
> > However, if we add a service that happens to have its users created
> > before a previously existing service, the UID's shift by one. When
> > this new image is deployed, the username might be 'ceilometer', but
> > /mnt/state/var/lib/ceilometer is now owned by 'cinder'.
> 
> Wow, nice find!
> 

Indeed, the Helion dev team discovered this one whilst working on updating
between alternating builds that added or removed some services.

Oh, I forgot to mention the bug reference:

https://bugs.launchpad.net/tripleo/+bug/1374626

> > 
> > Here are 3 approaches, which are not mutually exclusive to one another.
> > There are likely others, and I'd be interested in hearing your ideas.
> > 
> > * Static UID's for all state-preserving services. Basically we'd just
> >   allocate these UID's from a static pool and those are always the UIDs
> >   no matter what. This is the simplest solution, but does not help
> >   anybody who is already looking to update a TripleO cloud. Also, this
> >   would cause problems if TripleO wants to merge with any existing
> >   system that might also want to use similar UID's. This also provides
> >   no guard against non-static UID's storing things on the state
> >   partition.
> 
> +1 for this approach for the reasons mentioned.
> 
> > 
> > * Fix the UID's on image update. We can backup /etc/passwd and
> >   /etc/group to /mnt/state, and on bootup we can diff the two, and any
> >   UIDs that changed can be migrated. This could be very costly if the
> >   swift storage UID changed, with millions of files present on the
> >   system. This merge process is also not atomic and may not be
> >   reversible, so it is a bit scary to automate this.
> 
> If we really want to go with this type of aproach we could also just
> copy the existing /etc/passwd into the image thats being built. Then
> when users are added they should be added in after existing users.
> 

I do like this approach, and it isn't one I had considered. We will know
what image we want to update from in nearly every situation. Also this
supports another case, which is rolling back to the previous image,
quite well.

Really this is just an automated form of static UID assignment.