RE: Opsdir telechat review of draft-ietf-nvo3-hpvr2nve-cp-req-15

["Black, David" <[e-mail]>]

Scott,

Thanks for the review ... comments inline.

Thanks, --David

> -----Original Message-----
> From: Scott Bradner [mailto:[e-mail]]
> Sent: Saturday, February 17, 2018 8:20 PM
> To: [e-mail]
> Cc: [e-mail]; [e-mail]; draft-ietf-nvo3-hpvr2nve-cp-
> [e-mail]
> Subject: Opsdir telechat review of draft-ietf-nvo3-hpvr2nve-cp-req-15
> 
> Reviewer: Scott Bradner
> Review result: Has Nits
> 
> This is an OPS-DIR review of Split Network Virtualization Edge (Split-NVE)
> Control Plane Requirements (draft-ietf-nvo3-hpvr2nve-cp-req-15) This ID
> describes the requirements that should be met by the technology defined in
> one
> or more technical specifications to implement, in this case, control plane
> protocols for use in a split network virtualization system, and thus, by
> definition, can not have any direct impacts on the operation of networks.
> Technology specification(s) that meet these requirements might have such
> impacts. That said, some notes ---- section 1 .1 says The key words "MUST",
> "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT",
> "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as
> described in RFC 2119 [RFC2119] and RFC 8174 [RFC8174].
> 
> but RFC 8174 says the text should read as follows if the authors are following
> the guidance in RFC 8174
> 
> The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD",
> "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this
> document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when,
> and only when, they appear in all capitals, as shown here.
> 
> if the authors are not following that guidance  I would think that the original
> RFC 2119 wording should be used

The intent is to follow RFC 8174, where only upper case denotes an actual requirement,
in part because this document is providing requirements input to IEEE 802.

> ----
> the 3rd pp of section 2.2 says "
> The migrating VM should not be in Running state at the same time on the source
> hypervisor and destination hypervisor during migration
> 
> the pp goes on to say that it could happen

Not exactly ... Migration state occurs between the Running and Shutdown states
in both directions as described in the first paragraph of that section:

   The VM state on source hypervisor 1
   transits from Running to Migrating and then to Shutdown [RFC7666].
   The VM state on destination hypervisor 2 transits from Shutdown to
   Migrating and then Running.
 
> I may be missing something but this puzzles me a bit - I would think that the
> state of the results of the operation of the VM (e.g. database updates) would
> be indeterminable if both VMs are running at the same time without some sort of
> lockout -maybe this should be MUST NOT?

In practice, there is a lockout, but it only involves the hypervisors.  Both VMs are
in Migrating state during the transfer of execution from one host to another.
One of the benefits of this is that it reduces the number of entities that have to
participate in the transfer of control, as the states visible through the interface
change before and after transfer of execution, but not during that transfer.

Stating that both states are Migrating during transfer of execution could help
clarify this, and see below on removal of upper case RFC 2119 keywords from
all of section 3.
 
>  ----
> the 4th pp in section 3.1 states that "the external NVE MUST be notified ...
> when it no longer requires connection" - what happens if the software on the
> device needing the connection hangs - how is the notification done?

Something else has to decide that the device is dead and take action - that's
outside the scope of the protocol described here because another entity is
involved.  A sentence to note this need to clean up failed devices would make
sense to add.

> ---
> section 3 of the ID contains some things that appear to be requirements (i.e.,
> include MUST etc) and some things that appear to be descriptive - without
> calling out the actual requirements as is done in section 4 it will be hard for
> the technology specification developers to know what requirements they
> need to
> meet -if section 4 is a comprehensive list of requirements then it would seem
> to be a good idea to avoid 2119 type capitalization in section 3 - if not I
> would expect missed requiements

That sounds like the right approach, as the introductory text to Section 3
has this to say:

   The following subsections show illustrative examples of the state
   transitions of an external NVE which are relevant to Hypervisor-to-
   NVE Signaling protocol functionality. It should be noted this is not
   prescriptive text for the full state machine.