Re: [netmod] leafref to lists that contain system-controlled entries

[Robert Wilton <[e-mail]>]

Hi Jason,

Please see further comments inline ...


On 24/10/2017 00:58, Sterne, Jason (Nokia - CA/Ottawa) wrote:
>
> Thanks Rob. Please see below.
>
> Jason
>
> *From:*Robert Wilton [mailto:[e-mail]]
> *Sent:* Monday, October 16, 2017 6:40
> *To:* Sterne, Jason (Nokia - CA/Ottawa) <[e-mail]>; 
> [e-mail]
> *Subject:* Re: [netmod] leafref to lists that contain 
> system-controlled entries
>
> Hi Jason,
>
> On 13/10/2017 19:43, Sterne, Jason (Nokia - CA/Ottawa) wrote:
>
>     Hi all,
>
>     There are a few threads on the mailing list that touch on the
>     concept of system-controlled resources (mostly list entries):
>
>     https://mailarchive.ietf.org/arch/msg/netmod/3fTSHIh_MfHzmuDCoicAGiXA2E0
>
>     https://mailarchive.ietf.org/arch/msg/netmod/KIsSgKByQWpqYzA4i6Bwc8fuH3w
>
>     https://mailarchive.ietf.org/arch/msg/netmod/mjLJdiYErtNG41dJ5bJ5ji07cz0
>
>     A few drafts & RFCs also refer to the concept:
>
>     https://tools.ietf.org/html/draft-ietf-netmod-revised-datastores-04
>
>     https://tools.ietf.org/html/rfc7223
>
>     Several vendor implementations have list entries (instance data)
>     that are populated by the server and can be referenced (leafref)
>     from other places in the configuration.  These system entries are
>     useful pre-created policies, interfaces, etc that can then be used
>     (and referred-to) by operators in their explicit configuration.
>
>     If those entries are only expected to exist in the <operational>
>     datastore, then in theory any references to them in user created
>     configuration will cause a validation problem in the
>     candidate/running (missing leafref target).
>
>     One solution discussed in the mailing lists is to change every
>     reference to lists that could contain a system created entry to a
>     “require-instance false” leafref.  But then some useful validation
>     is lost. In many cases the model is more correctly
>     “require-instance true” but the set of targets includes the system
>     create entries.
>
>
> I agree that this is not a good general solution for system created 
> configuration that is always expected to exist on the device because 
> some of the useful validation is lost.
>
> But I think that this solution does work well were the system created 
> entries are truly dynamic in nature, e.g. it seems to work well for 
> the topology YANG module where the topology may be explicitly 
> configured, but would more normally be learned dynamically from 
> protocol interactions, or perhaps be configured via a dynamic 
> configuration protocol.
>
> *//*
>
> */[>>JTS] OK – I think I follow you here.  You’re saying that if there 
> are references to truly dynamic entries, then since those entries will 
> come and go (vs static bootup-time system entries that are always 
> there), references to them will more likely be “require-instance 
> false”.   But that does mean the system has to allow references to 
> non-existent entries, and you lose validation. You also risk errors 
> like referencing a name that is just 1 char different from what you 
> really wanted but the system can’t tell you that you got it wrong./*
>

Yes.  I don't think that you can solve this during existing YANG 
datastore validation, as it is defined today.

But I also think that NETCONF/YANG is potentially missing an RPC that is 
a step beyond validation, but before actually applying configuration.

YANG and the NMDA datastores draft makes it clear that both <running> 
and <intended> are always valid datastores. This means that the 
validation rules for these datastores really should not depend on the 
current hardware in a device if that hardware could be removed or 
change.  Otherwise, if you allow validation to depend on the current 
hardware capabilities, then if someone pulls out a linecard, that would 
cause a previously valid configuration to immediately become invalid, 
violating the rule that <running>/<intended> are always valid.

I think that a potential solution to this problem, is that a new NETCONF 
RPC could be defined that is a "<should-successfully-apply>" 
operation.   Processing during this new RPC would be able to check 
against current system resources, current hardware capabilities, etc, 
and would be designed to indicate whether the system expects (but does 
not guarantee) that the configuration would completely apply 
successfully without any errors or unapplied configuration.  
Configuration datastores would not have to always conform to this 
constraint, so that if an operator changed or removed a linecard, the 
configuration would still be "valid" (as per NMDA and RFC7950 rules) but 
would fail a subsequent "should-successfully-apply" check.



> *//*
>
>     Another solution discussed is to have the system created entries
>     appear in the <intended> datastore (as part of
>     template/expansion).  That would make validation pass on the
>     intended datastore, but then the candidate/running/startup
>     datastores would not be valid (would be missing leafref targets if
>     any part of the config refers to system created entries).  THis
>     sounds similar to the problem that has been discussed in the past
>     about the fact that templates (in the running) basically mean the
>     running/candidate aren’t necessarily valid (until after template
>     expansion, which means only the intended would be valid).
>
>
> I think that this solution is OK, but not necessarily ideal.
>
> As you say, it means that <running>/<candidate>/<startup> may not be 
> valid, which I see as quite a big down side.  Longer term, I think 
> that it would be a good aim to allow the configuration to be validated 
> off the device, if a client desired to do so.
>
>
>
>     Another approach could be to actually have those system created
>     entries show up in running/candidate.  That would ensure that
>     references to those entries are valid.  But if the whole concept
>     of templates just cause the running/candidate to not be valid
>     anyways maybe we wouldn’t worry about the invalid aspect of
>     references to system created list entries ?
>
>
> I know that quite a few implementations do this today, but I'm 
> generally not a fan of the system modifying <running>.  It seems that 
> an overall architecture is much cleaner if <running> has a single 
> source of truth and hence can be exclusively controlled by the client.
>
> But I also like the approach where the client (rather than the device) 
> explicitly writes these default entries into the configuration, if 
> they are referenced elsewhere by the configuration, to make the 
> configuration "complete".  E.g. if part of the configuration 
> references the loopback0 interface then also explicitly add the 
> necessary loopback0 configuration to instantiate the "loopback0" 
> interface. When this configuration is pushed to the device (i.e. using 
> merge or replace operation semantics) then the system should silently 
> accept/ignore the explicit configuration to create the loopback0 
> interface if it already exists on the system.
>
> *//*
>
> */[>>JTS] Yes – that is another option and I like it.  If I follow you 
> correctly, the server would never return these system entries in a 
> <get-config> unless the client/operator had already explicitly 
> ‘created’ them. /*
>
> */So the operator has the option to make the system entries visible or 
> not./*
>
Yes, exactly.

/*
*/
>
> *//*
>
> *//*
>
> */I think the server should still accept references to the system 
> entries even if the client/operator hasn’t explicitly created them.  
> The whole point is that those system entries are there and waiting for 
> operators to use them from the start (without *having* to explicitly 
> create or define them).   In that case the references would be 
> ‘dangling’ (unresolved) as far as an offline validation is concerned 
> (but a client could select to fix that by explicitly defining any 
> entries they want to reference)./*
>
I think that this is OK.

Effectively, I see that being like a static system provided template for 
configuration that is merged with <running> to form <intended>, which is 
then validated.


> *//*
>
>
>
>
> At the moment, IETF, and other SDOs are busy defining standard YANG 
> models, but for those models to end up being truly generic they also 
> need to have consistency about which bits of configuration are always 
> expected to implicitly exist on the device.  E.g. considering the 
> example above of configuration referencing loopback0: if some systems 
> automatically create a loopback0 interface and others do not, then a 
> generic configuration needs to handle both scenarios.
>
> If IETF standardizes YANG configuration templates, then perhaps it 
> would be good to investigate whether some of these "useful default 
> system properties" could instead be embodied into one or more standard 
> device templates?  These templates could then be explicitly referenced 
> in the <running> configuration.  This may allow <running> to be small, 
> but still allow it to be "complete" and able to be validated off the box.
>
> */[>>JTS] I’m not clear on this approach. /*
>
OK, so this is just an idea:

1) Assume a YANG extension is defined to allow templates to be defined.
2) Further, assume that there is a way to store, and uniquely name some 
of those templates in a standard place.

So, perhaps IETF could define a template like this, which is stored as a 
well defined place:

<template>
   <name>ietf-basic-router-v1</name>
   <config>
     <interfaces>
       <interface>
         <name>Loopback0</name>
         <type>ianaift:loopback</type>
       </interface>
       <interface>
         <name>Null0</name>
         <type>ianaift:null</type>
       </interface>
     </interfaces>
   </config>
</template>

Now, when it comes to the configuration file for your device, it could 
look like this:
   <config>
<use-remote-template>http://yang-templates.ietf.org/ietf-basic-router-v1</use-remote-template>

     ... rest of config as normal.
   </config>

So, the expanded configuration would include the explicit configuration 
for Loopback0 and Null0 interfaces, but that would be pulled via use of 
a remote template (the contents of which is probably already cached on 
the device).


> *//*
>
> */How would the templates be referenced in the <running> ?  Can you 
> give me an example ? (is this different than a direct reference to a 
> system created entry that I am talking about ?)/*
>
The above is just an idea.  Normally I would expect configuration 
templates to be defined in the running configuration.  But here I was 
considering the idea that a template is predefined in some way, and then 
referenced, so that it doesn't have to be provided inline in the running 
configuration.


> *//*
>
> */How would this allow <running> to be small ? Do the templates 
> contain the full definition of the system created entries ?/*
>
Yes, I was assuming that the template could contain what might normally 
be represented as system created entries today.  Standard templates 
could either be defined by SDOs, vendors, or operators, as long as there 
is a way to reference them.

Thanks,
Rob


> *//*
>
>
>
> Thanks,
> Rob
>
>
>
>     Rgds,
>
>     Jason
>
>
>
>
>     _______________________________________________
>
>     netmod mailing list
>
>     [e-mail] <mailto:[e-mail]>
>
>     https://www.ietf.org/mailman/listinfo/netmod
>