heat stack-update and scale-in/scale-out

[alonsocamaro]

In general we need that our heat resource types support the stack-update operation. This is needed to fix failed stacks otherwise they have to be left stale in the system. I understand updating all the resource types requires a major effort.

On the other hand, stack-update is also needed to allow for scaling-in/scaling-out services (typically changing the number of pool members). This is a must have operation which we cannot do currently.

From my point of view allowing stack-update operations on the F5::Sys::iAppService resource type should be on top of our list.

Regards

   Ulises

[dflanigan]

@pjbreaux could you scope the work needed for these two issues: #126 and #127

[pjbreaux]

This work would take around a week.

[alonsocamaro]

I understand/guess that with orthogonal you mean regular scale-in/out should be implemented like https://github.com/F5Networks/f5-openstack-heat/issues/130 ?

Just wanted to comment that when the F5 itself is the only element implementing the functionality (ie: no servers used) like in the case of APM (ie: number or users), PEM/AFM (ie: (ie:bandwidth allocation), only stack-update would be needed to update to provide full scale-in scale-out. The feature requested in this entry would also suffice the scale-in/scale-out scenarios when this scale-in/out is triggered by an externally (ie: unlike https://github.com/F5Networks/f5-openstack-heat/issues/130) to the scaled components.

Also, last but not least. stack-update would also be needed to change the configuration in general (ie: not scale-in/scale-out) of a deployed iApp.

[alonsocamaro]

In summary I find we need this to:

• fix failed stacks
• reconfigure the stack (ie: updating any setting)
• scale-in/scale-out (ie: adding/removing members) which could be seen as a subset of reconfiguration use case

[davmec]

+1 This fix would be a must to use F5 heat templates for NFV Virtual Gi LAN deployments. Capacity grows continuously in such scenarios and the option to update existing multi-VE HEAT deployments (scale-out) with additional VEs is very important for SP business continuity.