Open Stack

Excerpts from Ton Ngo's message of 2014-07-29 13:53:12 -0700:
> 
> Hi everyone,
>     The raw template saved in the DB used to be the original template that
> a user submits.  With the recent fix for stack update, it now reflects the
> template that is actually deployed, so it may be different from the
> original template because some resources may fail to deploy.  I would like
> to solicit some feedback on saving the original template in the DB
> separately from the deployed template.  I can think of two use cases for
> retrieving the original template:
>    Debugging:  running stack-update using the same template after fixing
>    environmental problems.  The CLI and API can be extended to allow
>    reusing the original template without having to provide it again.
>    Convergence or retry:  some initial resource deployment may fail
>    intermittently, but the user can retry later.
> 

I believe this use case is far better handled via vcs. We need the
template to parse the current state of the stack. The user will have
their intended template and can have their intended parameter values
all included in a VCS.

>      Are there other potential use cases?    The cost would be an extra
> copy of the template in the raw template table for each stack if there is
> failure, and a new column in the stack table to hold the id.  We can argue
> that the user should have the original template to resubmit, but it seems
> useful and convenient to save it in the DB.
> Ton Ngo,
> 

Additional cost is the additional complexity of code to manage the data.