Sprint Feedback

Tom Haddon tom.haddon at canonical.com
Tue Jun 7 14:31:11 UTC 2011


On Tue, 2011-06-07 at 11:05 -0300, Gustavo Niemeyer wrote:
> > Sure, we're all abstracting things out and trusting other people with
> > certain parts of the system. I'm not writing all the code that runs on
> > my system, I'm using packaged code, etc.,
> 
> Right, exactly.
> 
> > but that doesn't mean you still don't have visibility on the changes
> > being made or having control over what changes you're comfortable
> > with.
> 
> Interestingly, it does mean as much.  Last time you upgraded the
> kernel, have you inspected the code to see what were the actual
> changes?  No, because that's too time consuming, too complex
> to observe and understand it all at once, and because there are
> good people taking care of that.  So, you choose to delegate
> control over your production environment to these people.

I absolutely take the point that we're not looking at all the kernel
code every time we upgrade the kernel. However, what I meant was more
that I do have a reasonable expectation that the configuration options I
have configured for kernel settings won't get overwritten with each new
upgrade. So yes, I am trusting in the sense of not knowing about
specific changes, but I am also controlling in the sense of certain
exposed configuration options (hence the term, configuration
management).

> Ensemble is way too young for me to request that level of trust
> from you, of course.  What I'm pointing out, though, is that the.
> idea that people want to know _everything_ going on is simply
> not true.  

Absolutely. I'm not saying we know everything going on.

> Ensemble adds a layer of abstraction, and like all
> abstraction layers, it hides some details.  Even then, if it
> works well, and makes people's lives significantly easier, and
> enables them to do things they couldn't otherwise, it will
> succeed.

Yep, agreed. Adoption will also depend (to some extent) on how
comfortable people are with the level of exposure to the details that
Ensemble gives. If it hides too much, that'd be a problem, just as if it
didn't hide enough, that'd be a problem :)

> > So fwiw, we can (and are) doing much of this in Puppet as well. I won't
> > say it's as easy syntactically as Ensemble promises to be, but we do
> > have services that we can scale up very trivially and other parts of the
> 
> Sure, in essence that's doable even with shell scripts.  It's
> analogous to the way we also don't need packages or apt: they're just
> a bunch of files in a tarball, and you can also do it by hand, or code
> your own system if you want, as many have been doing.  Rolling your
> own has good parts (e.g. you understand it all) and bad parts (e.g.
> you maintain it all, and don't benefit of work others have done that
> could otherwise be plugged in the system in an effortless way).

I'm not really sure I follow here. Are you saying that you could in
theory do what Puppet is doing with shell scripts? Surely that applies
just as much to any configuration/deployment tool?





More information about the Ensemble mailing list