Re: [Netconf] issues with processing onboarding information (zerotouch)

Hi,

Kent Watsen <kwatsen@juniper.net> wrote:
> Dear WG,
> 
> Following up on this thread, I'm beginning to make the changes,
> but there are a couple items where input would be helpful:
> 
> 1) regarding this:
> 
>    > b) Clarify that the scripts MUST also be idempotent, in case
>    >    the bootstrapping process falls into a loop.  Alternatively,
>    >    we could introduce a requirement on the scripts to supply
>    >    some sort of clean-up command; then the only state retained
>    >    would ever be the currently running boot-image, which is 
>    >    fine.  Thoughts?
> 
> What's being discussed here is the case where the script succeeds, 
> but a subsequent step (e.g., commit) fails.  In this case, do we
> say:
> 
>  a) scripts MUST be idempotent.  This sounds good, but I wonder
>     how possible this is.
> 
>  b) scripts MUST supply a clean-up command that removes all state
>     created by a successful execution of the script.  This seems
>     easier to code.

Do you mean that each script takes some kind of (conceptual) input
parameter that tells it to clean up the state produced by the previous
invocation?

Could we instead leave this to implementations?  In some environments
the server might be able to do the cleanup itself (which would be less
error prone).  In some other environments the scripts might not even
create any state.

/martin

> 
> Note: if no objections, I'll go with (b).
> 
> 
> 
> 2) regarding this:
> 
> 
>    > Should we add more zerotouch "progress-type" enums, and maybe
>    > make more of them mandatory?
>    >
>    > Details: module ietf-zerotouch-bootstrap-server contains the 
>    > RPC report-progress, which has input leaf "progress-type",
>    > which is an enumeration.  Currently, the enums follow
>    > this pattern:
>    >
>    > - bootstrap-initiated
>    > - bootstrap-complete
>    > - <step>-warning
>    > - <step>-error
>    > - informational
>    >
>    > where <step> has values: parsing, boot-image, pre-script,
>    > config, and post-script.
>    > 
>    > a) Should we add additional well-typed values for visibility
>    >    reasons (i.e. more debug information sent to the bootstrap
>    >    server)?  Specifically, these two:
>    >
>    >     - <step>-initiated
>    >     - <step>-success
>    > 
>    > b) assuming (a), should we make more of the reporting of
>    >  progress mandatory?  Currently only "bootstrap-complete"
>    >  is mandatory, with everything else being a SHOULD.
> 
> That kind of says it all, any thoughts on (a) and (b)?
> 
> Note: if no objections, I'll go with (a) only and, for (b), just
> make "bootstrap-initiated" mandatory.
> 
> 
> 
> Thanks,
> Kent // author
> 
> 
> 
> 
> _______________________________________________
> Netconf mailing list
> Netconf@ietf.org
> https://www.ietf.org/mailman/listinfo/netconf
> 