----- Original Message -----
> On Apr 24, 2011, at 8:27 PM, R.I.Pienaar wrote:
> 
> > 
> > On 24 Apr 2011, at 17:24, Luke Kanies <[email protected]> wrote:
> > 
> >> On Apr 22, 2011, at 11:44 AM, Daniel Pittman wrote:
> >> 
> >>> On Fri, Apr 22, 2011 at 10:18, Arm Adam
> >>> <[email protected]> wrote:
> >>> 
> >>>> Does puppet have any concept of a job engine with a request ID
> >>>> so that
> >>>> I can submit a request for a kick to happen, be given a request
> >>>> ID,
> >>>> and then be able to check back in for results for that specific
> >>>> request ID?  Or is the only way of knowing when it is done to
> >>>> run kick
> >>>> in --foreground mode?
> >>> 
> >>> No.  Specifically, Puppet doesn't even come close to having a
> >>> model of
> >>> that for the network.  What we do have is MCollective, which is
> >>> much
> >>> closer.  I don't recall if it does asynchronous operations yet,
> >>> but it
> >>> certainly allows the sort of remote control and verification on
> >>> puppet
> >>> runs that you are after.
> >> 
> >> As Daniel says, mcollective is the only real way to do this kind
> >> of work right now, but it is still essentially syncronous in
> >> terms of control - you can't set up jobs and manage them
> >> discreetly, as far as I know.
> >> 
> >> However, I've been thinking a lot about this recently, and I
> >> expect we'll be looking at whether it's reasonable to add
> >> something like it to the system at some point, so I'd love to
> >> hear about anything you come up with.
> >> 
> > 
> > 
> > This kind of use is spec'd on the mcollectiv road map and there is
> > some work going on at the moment to make this a good fit - mixing
> > real time vs scheduled jobs.
> 
> That's good to hear.  I'm also interested in what amounts to
> background jobs - something you start immediately, but isn't run
> with a console attached.  I don't know if this is any different in
> infrastructure from scheduled jobs, though.

It can do this already - today they run immediately in future you can
schedule them to run at 4am or whatever:

% mco rpc --nr package yum_clean
Request sent with id: 405d5413f423ed1b370539ac8ef9c7c2

and if we used time to measure how long this took from the user point of
view:

mco rpc --nr package yum_clean  0.30s user 0.03s system 91% cpu 0.357 total

So all the machines are clearing their caches, I dont particularly care
for the success or fail of this so dont want the interactive feedback and
dont want to waste my time waiting for useless feedback.  I use this for 
example to clear my yum caches whenever I add a package to the yum repo

The eventual plan is that should you later start caring for the output
you can ask it the result of a given job - 405d5413f423ed1b370539ac8ef9c7c2 
in this case - and you can gain access to RPC results same as you do now


> 
> > There is a prototype using puppet dsl as a way to express jobs and
> > relations between them with failure resolution etc but this will
> > work better further down the line
> > 
> > I plan to work on this during the 1.3 x development series that
> > should kick off soon, if you look in the mcollective issues list
> > and roadmap on docs.puppetlabs.com you will see some more details
> > on the current plans
> > 
> > I am actively seeking feedback and requirements though to help plan
> > this work as surprisingly it is not a feature that has actually
> > been requested much so the picture of what people need isn't
> > extremely clear.
> 
> I know as we look at adding support in our web tools for phased
> roll-out, we'll need this ability, but we haven't had enough time to
> actually spec it out yet.

Yup, I'd like mcollective to be an appropriate transport for BPM systems
so will need that and a lot more.

-- 
You received this message because you are subscribed to the Google Groups 
"Puppet Developers" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/puppet-dev?hl=en.

Reply via email to