Open Stack

On Fri, Feb 28, 2014 at 03:45:39PM -0500, Zane Bitter wrote:
> On 28/02/14 02:28, Qiming Teng wrote:
> >
> >The creation a stack is usually a time costly process, considering that
> >there are cases where software packages need to be installed and
> >configured.
> >
> >There are also cases where a stack consists of more than one VM instance
> >and the dependency between instances.  The instances may have to be
> >created one by one.
> >
> >Are Heat people considering adding some progress updates during the
> >deployment?  For example, a simple log that can be printed by heatclient
> >telling the user what progress has been made:
> >
> >Refreshing known resources types
> >Receiving template <...>
> >Validating template <...>
> >Creating resource my_lb [AWS::EC2:LoadBalancer]
> >Creating resource lb_instance1 [AWS::EC2::Instance]
> >Creating resource latency_watcher [AWS::CloudWatch::Alarm]
> >....
> >...
> >
> >
> >This would be useful for users to 'debug' their templates, especially
> >when the template syntax is okay but its activities are not the intended
> >one.
> 
> Yes, we need some sort of back-channel to feed information to the
> user - not only on progress but things like warnings. Right now we
> have to choose between failing a whole stack or not notifying the
> user at all when something is suspect.
> 
> The ReST model is unfortunately not conducive to this, as you
> generally don't want to keep the HTTP connection open and block
> until something is complete. One good idea floating around is to
> send the messages to a Marconi queue that the user can connect to.
> 
> It's all up for discussion in this blueprint:
> 
> https://blueprints.launchpad.net/heat/+spec/user-visible-logs
> 
> >Do we have to rely on heat-cfn-api to get these notifications?
> 
> No, you can also use the native api ;)
> 
> cheers,
> Zane.
> 

Thanks. 'User-visible-logs' is indeed what I am really interested in.

Better yet, the in-instance log can be fed back to users.

Regards,
  Qiming