Re: [lmap] default value for the controller-timeout

["MORTON, ALFRED C (AL)" <[e-mail]>]

> -----Original Message-----
> From: Juergen Schoenwaelder [mailto:j.schoenwaelder@jacobs-
> university.de]
> Sent: Thursday, February 04, 2016 10:41 AM
> To: MORTON, ALFRED C (AL)
> Cc: [e-mail]
> Subject: Re: [lmap] default value for the controller-timeout
> 
> On Thu, Feb 04, 2016 at 09:42:27AM -0500, MORTON, ALFRED C (AL) wrote:
> > > -----Original Message-----
> > > From: Juergen Schoenwaelder [mailto:j.schoenwaelder@jacobs-
> > > university.de]
> > > Sent: Thursday, February 04, 2016 9:15 AM
> > > To: MORTON, ALFRED C (AL)
> > > Cc: [e-mail]
> > > Subject: Re: [lmap] default value for the controller-timeout
> > >
> > > On Thu, Feb 04, 2016 at 08:53:18AM -0500, MORTON, ALFRED C (AL)
> wrote:
> > > > > -----Original Message-----
> > > > > From: Juergen Schoenwaelder [mailto:j.schoenwaelder@jacobs-
> > > > > university.de]
> > > > > Sent: Wednesday, February 03, 2016 11:59 AM
> > > > > To: MORTON, ALFRED C (AL)
> > > > > Cc: [e-mail]
> > > > > Subject: Re: [lmap] default value for the controller-timeout
> > > > >
> > > > > On Wed, Feb 03, 2016 at 08:57:56AM -0500, MORTON, ALFRED C (AL)
> > > wrote:
> > > > > > > -----Original Message-----
> > > > > > > From: lmap [mailto:[e-mail]] On Behalf Of
> Juergen
> > > > > > > Schoenwaelder
> > > > > > > Sent: Wednesday, February 03, 2016 7:09 AM
> > > > > > > To: [e-mail]
> > > > > > > Subject: [lmap] default value for the controller-timeout
> > > > > > >
> > > > > > > Hi,
> > > > > > >
> > > > > > > should the ma-config-controller-timeout (information model)
> have
> > > a
> > > > > > > default value? The YANG model has this as an optional
> object.
> > > What
> > > > > > > does it mean if lmap/agent/controller-timeout is not
> present? I
> > > > > think
> > > > > > > I would prefer to have a default so that there is always a
> > > timeout.
> > > > > > >
> > > > > > > What about a week (= 604800 seconds) as a default?
> > > > > > >
> > > > > >
> > > > > > [ACM] Good to supply a default, 1 week ok, not sure if timeout
> > > should
> > > > > be
> > > > > > optional in the YANG model (even if the timeout object is
> omitted,
> > > a
> > > > > > timeout is enforced at the MA?
> > > > > > How to instruct the MA not to use a timeout?)
> > > > > >
> > > > >
> > > > > While coding this up I stumbled over this. Right now, the text
> is
> > > not
> > > > > very precise. I could go with setting the timeout to zero means
> no
> > > > > timeout and the default is one week.
> > > > >
> > > > > /js
> > > > >
> > > >
> > > > [ACM]
> > > > As an alternate, we could view the default simply as guidance.
> > > > If you include the ma-config-controller-timeout object, then
> > > > you must supply a value, and if you need guidance about this you
> > > > specify the default value.
> > > >
> > > > OTOH, if you don't want a timeout, you don't include the
> > > > ma-config-controller-timeout object at all.
> > > >
> > > > AND, I don't feel strongly about this...
> > >
> > >
> > > Yes, but this means the default is actually to have no timeout since
> > > the ma-config-controller-timeout is optional to have. It's a policy
> > > question whether we want to have a timeout configured by default and
> > > it takes effort to run without one or whether we are fine with no
> > > timeout by default and we require effort to create one.
> > >
> >
> > [ACM] Right. The default condition of the MA is having no instructions
> > and having nothing scheduled, etc.  When we ask the MA to perform
> > measurements that send traffic, reports, etc., then we may need a
> timeout,
> > and all these functions require effort to create them.
> >
> 
> We are talking about loosing contact to the controller, which should
> trigger an appropriate event. Relying on someone to configure this in
> a meaningful way (= having no default) means we will see boxes that
> simply will never react to lost contact to a controller. Yes, it is a
> policy issue but I am somewhat concerned about boxes going crazy and
> never detecting this and properly shutting down.
> 

[ACM] My concern as we started this discussion is about having 
an "embedded behavior" = default timeout in the MA.  
I'm using this term "embedded behavior" for lack of a better one,
since the MA would apparently need to be instantiated with this timeout.
The framework mentions the timeout configuration (end of 5.3.1):
http://tools.ietf.org/html/rfc7594#section-5.3.1

   The Controller may want an MA to run the same Measurement Task
   indefinitely (for example, "run the 'upload speed' Measurement Task
   once an hour until further notice").  To prevent the MA continuously
   generating traffic after a Controller has permanently failed (or
   communications with the Controller have failed), the MA can be
   configured with a time limit; if the MA doesn't hear from the
   Controller for this length of time, then it stops operating
   Measurement Tasks.

but doesn't mention an embedded default for this time limit.

After having read 
https://tools.ietf.org/html/draft-ietf-netmod-opstate-reqs-04
yesterday (which I know you've read, too)
the "embedded behavior" = default timeout in the MA
isn't part of "Applied Configuration" or "Derived State",
so it may not be possible to retrieve the timeout setting.

What is the down-side to making the 
ma-config-controller-timeout object Mandatory? 
Having to repeat the object with every Instruction?
But this would make the timeout part of the retrievable configuration,
and that's what I would like to see, if it's not a burden elsewhere.

Al