fuel-dev team mailing list archive

[Mike Scherbakov <mscherbakov@xxxxxxxxxxxx>]

+fuel-dev


On Thu, Jan 30, 2014 at 8:45 PM, David Easter <deaster@xxxxxxxxxxxx> wrote:

> The requirements were changed after the discussion with Evgeniy to reduce
> the scope/concerns about the feature.
>
> The primary use cases to address are now:
>
> 1. The user is deploying from scratch and realizes a major mistake in
> choices has been made.  In this case, the deployment should be reset to
> its initial state and restarted completely.
>
> 2. The user is deploying from scratch and a major failure occurs - e.g.
> one or more controllers fail to deploy.  In this case, the deployment
> should be reset to its initial state and restarted completely.
>
> 3. The user are doing an incremental deployment and adding a new compute
> (ceph, etc) node to the cluster, and those actions are failing.  In this
> case, only the incrementally added nodes should be reset to initial state
> and when the deployment is stopped and only those nodes would be
> redeployed.
>
> 4. The user deletes nodes from an existing deployment.  In this case, the
> action cannot be stopped - it will always complete and return the nodes to
> their initial state.
>
> So the concerns about intermediate state are removed.  The only time you
> won¹t restart the entire deployment is where there was a working
> environment to begin with and all you¹re doing is adding new nodes.
>
>
> Reset Environment is *only* for environments that are already deployed.
>
> Additional conversation happened in Skype.  Just sending this for those
> that weren¹t on the chat.
>
> Thanks,
>
> -Dave Easter
>
>
>
> On 1/30/14, 7:37 AM, "Nikolay Markov" <nmarkov@xxxxxxxxxxxx> wrote:
>
> >Hello David,
> >
> >There may be some inaccuracies in how we understand these two
> >features, so I decided to clarify some things.
> >
> >1) Stop deployment
> >As I see from your discussion with Evgeniy, it should work like that:
> >User may click "Stop" button during deployment running to interrupt it
> >and then get into some kind of "intermediate state". In this state he
> >can't modify any settings, but can do one of two things - either click
> >"Reset" button (2) or choose someting like "Redeploy error nodes" or
> >just "Continue".
> >
> >First case is then one or several errors occured during deployment and
> >user doesn't want to wait till it ends, he just interrupts it and
> >environment reverts to exact state that was before clicking "Deploy
> >changes", allowing to tune some options again. By the way, all this
> >may also be done with one "Reset" button.
> >
> >Second case is then some nodes turn to error state during deployment,
> >some are deployed successfully, but the deployment is still running
> >and isn't completed. So, user may stop the deployment (settings will
> >remain locked), do something (like, turning back on node which gone
> >offline) and then press "Continue", so Fuel will reprovision and
> >redeploy ONLY nodes which weren't ready then user clicked "Stop"
> >button.
> >
> >2) Reset environment
> >As a said above, user may reset an environment during deployment to
> >revert it completely to starting stage. But also it will work the same
> >on an environment which is already deployed, rebooting nodes back to
> >bootstrap and allowing to run the whole process for whole cluster
> >again.
> >
> >
> >Please correct me if I misunderstood something, so we could get rid of
> >any issues and delays.
> >
> >
> >--
> >Best regards,
> >Nick Markov
>
>
>


-- 
Mike Scherbakov
#mihgen