Re: [VRRP] draft-ietf-vrrp-unified-mib-07.txt MIB Dr. Review

[Mukesh Gupta <[e-mail]>]

WG Members,

Please voice your opinions about this so that we can move forward with the MIB draft.

Kalyan/Joan, historically, we have not heard a lot of opinions on the MIB draft on the WG mailing list, probably because we do not have enough MIB experts in the WG.  So, we have relied on the feedback/opinion of the MIB doctors.

[speaking as a WG member]
I am not a MIB expert myself but, after reading through your discussion, it does seem to me that a new clean MIB would be a way to go.

Regards
Mukesh

-----Original Message-----
From: [e-mail] [mailto:[e-mail]] On Behalf Of Kalyan (Srinivas)Tata
Sent: Tuesday, March 16, 2010 11:51 AM
To: Joan Cucchiara; Romascanu, Dan (Dan); Adrian Farrel; Ronald Bonica; [e-mail]; [e-mail]
Cc: MIB Doctors (E-mail)
Subject: Re: [VRRP] draft-ietf-vrrp-unified-mib-07.txt MIB Dr. Review

Hi Radia and Mukesh,
        I think we need WG to express opinion on whether to start with a new MIB or continue with this one. This will determine how to go about with changes suggested by Joan.

Thanks,
Kalyan

-----Original Message-----
From: Joan Cucchiara [mailto:[e-mail]]
Sent: Wednesday, March 10, 2010 10:55 AM
To: Kalyan (Srinivas)Tata; Romascanu, Dan (Dan); Adrian Farrel; Ronald Bonica; [e-mail]; [e-mail]
Cc: MIB Doctors (E-mail)
Subject: Re: [VRRP] draft-ietf-vrrp-unified-mib-07.txt MIB Dr. Review


Hi Kalyan,

Thank you for your quick response.  Please see
comments inline.

Thanks,
  -Joan


> ----- Original Message -----
> From: "Kalyan (Srinivas)Tata" <[e-mail]>
> To: "Joan Cucchiara" <[e-mail]>; "Mukesh Gupta"
> <[e-mail]>; "Romascanu, Dan (Dan)" <[e-mail]>; "Adrian
> Farrel" <[e-mail]>; "Ronald Bonica" <[e-mail]>;
> <[e-mail]>; <[e-mail]>
> Cc: "MIB Doctors (E-mail)" <[e-mail]>; "Radia Perlman"
> <[e-mail]>; <[e-mail]>;
> <[e-mail]>
> Sent: Tuesday, March 09, 2010 2:27 AM
> Subject: RE: [VRRP] draft-ietf-vrrp-unified-mib-07.txt MIB Dr. Review
>
>
> HI Joan,
>
> My comments inline.
>
>>
>> MIB Compiler: SMICng
>> --------------------
>>
>> W: f(VRRP-MIB.my), (437,8) Row "vrrpAssociatedIpAddrEntry" has indexing
>>   that may create variables with more than 128 sub-ids
>> W: f(VRRP-MIB.my), (158,23) Row "vrrpOperationsEntry" does not have a
>>   consistent indexing scheme - index items from current
>>   table must come after index items from other tables
>> W: f(VRRP-MIB.my), (453,23) Row "vrrpAssociatedIpAddrEntry" does not have
>>   a consistent indexing scheme - cannot specify an index item from
>>   additional "base row" vrrpOperationsEntry, since can have
>>   only one "base row" which is ifEntry
>>
>>
>> WRT the INDEX To Large error.  I think this was addressed by Bert Wijnen
>> in
>> prior
>> versions but I don't see any wording in the DESCRIPTION clause or any
>> SIZE
>> limit
>> on the Index vrrpAssociatedIpAddr.  So something like this (and some text
>> describing
>> why the size limits in the DESCRIPTION clause would be appropriate.
>> Additionally,
>> the vrrpOperationsInetAddrType should be restricted.
>>
>>
>>       vrrpAssociatedIpAddr OBJECT-TYPE
>>           SYNTAX       InetAddress (SIZE(0|4|16))
>>           MAX-ACCESS   not-accessible
>>           STATUS       current
>>           DESCRIPTION
>>
>
> [Kalyan>] Will do.
>
>> Please discuss the other 2 warnings wrt Indexing.  Specifically, why are
>> tables using AddrType, VrId and ifIndex, rather than ifIndex, VrId,
>> AddrType?
>> (Please see comment below regarding section 8.)
>>
> [Kalyan>] The table indexing order was suggested by Bert during very
> initial review of the draft. The comment was that it would be easier for
> the administrator to sort based on the address type. Though I am not sure
> if I changed from ifIndex, VrId, AddrType. I probably had it wrong at that
> time too. I will change this ifIndex, VrId, AddrType
>

Okay, I can see his point that accessing by IP addresses has some
advantages.   Unfortunately,  the way the tables are currently, you'd
need to have "AddrType, Addr"   together in order to do this.

Let's leave this an open issue for the moment and close on the other issues
first.


>>
>> MIB Compiler: SmiLint
>> ------------------------
>> mibs/VRRP-MIB.my:437: [5] {index-exceeds-too-large}
>>   warning: index of row `vrrpAssociatedIpAddrEntry'
>>   can exceed OID size limit by 142 subidentifier(s)
>>
>>
>> Comments
>> --------
>>
>> 1)NIT:  Does not strip cleanly with SmicngPRO mstrip
>> (This doesn't need to be fixed for MIB Dr. review, just
>> as an fyi).
>
> [Kalyan>] Will fix it.
>
>> 2) Please Discuss:
>> My understanding (based on an old email exchange (fall 2008) with the MIB
>> author
>> and the AD of the WG at that time) was that the VRRP unified MIB
>> was not going to be an update of the original VRRP-MIB (RFC2787), but
>> rather a new MIB based on the new unified spec that was being
>> created.  There are a couple issues that I'd like clarified wrt to an
>> updated MIB (this draft) vs. a new MIB:
>>
>>
>> A) draft-ietf-vrrp-unified-spec-05.txt, section 8.4 states "that VRRPv2
>> and
>> VRRPv3
>> interoperation is optional" and the
>> recommendation is that it should only be done for transitioning from
>> VRRPv2 to VRRPv3.  A new MIB (rather than an updated
>> MIB) would likely re-enforce the guidelines from the specification.
>> Could you discuss why an updated MIB is better in this regard?
>>
>>
>> B)  Additionally, this new MIB is based on a MIB that is 10 years old
>> (RFC2787),
>> while not specifically an issue, I do see that this updated MIB is
>> using older MIB conventions.  Not a show-stopper but certainly a
>> consideration.  Was this considered?
>>
>> C) Comment: Most of the MIB objects contained in RFC2787 are deprecated
>> by this MIB. This indicates that creating a new MIB would be
>> fairly straightforward.
>>
> [Kalyan>] We had this discussion during the first draft. At that time
> updating made more sense but now with a separate unified spec, probably
> creating a new MIB would be better. Also my reasoning was that the
> previous draft has gone through multiple reviews and revisions, I wanted
> to just make the required changes to quickly get this done with less
> issues (and less work from me :) ).  But, I see your point, a new draft
> would certainly be much cleaner and probably the way to go.



Would be good for the WG to give an opinion on this issue since LC
is going on now.

>
>> 3) The terminology of this document does not consistently utilize the
>> Definitions
>> outlined in the draft-ietf-vrrp-unified-spec-05.txt.  So for example,
>> VRRP is used in places to mean VRRP Router.  The use of IPvX is not
>> done.  Would be better to be consistent between the spec and the MIB.
>>
> [Kalyan>] Will be more aware of this in the next revision.
>
>
>> 4)  DEPRECATED OBJECT DESCRIPTIONS:
>>
>> Way back when, Dave Perkins suggested starting the DESCRIPTION clause
>> for a deprecated object with:
>>
>> "This object is DEPRECATED.  This is done because..."
>> followed by the original description.
>>
>> While you have done a great job including the reasons why an object was
>> deprecated, could you move this to the beginning of the DESCRIPTION
>> clause
>> (rather than end)?
>>
>>
> [Kalyan>] Will do.
>
>> 5) Title: Please mention that this is Version 3 somewhere in the title
>> (as does the specification)
>>
> [Kalyan>] I remember asking Mukesh if we should change the title and I
> think he mentioned that we keep the same one. We can change the title if
> we are going with a new MIB.
>

Agreed.


>> 6) Section 4. Relationship to RFC2787
>>
>> The following sentence is troublesome because the MIB Module "deprecates"
>> objects and does not obsolete them.  While you may be trying to indicate
>> that this document will obsolete RFC2787, that is not exactly clear.
>>
>> (perhaps, this is another reason to create a new MIB module vs. an
>> updated
>> one?)
>>
>>   "RFC2787 [RFC2787] defines managed objects for VRRP over IPv4 and is
>>   now obsoleted by this memo."
>>
> [Kalyan>] I was meaning to say that this document will obsolete RFC2787. I
> get it :), going to a new MIB probably is the right thing.
>
>> 7) Section 5. Relation to Interface Group (IF-MIB)
>>
>> A) Please remove the word "physical" in front of physical interface.
> [Kalyan>] Will do.
>
>> B) Should specify a reference such as:  The VRRP-MIB Module imports
>> ifIndex from the IF-MIB.  At this time, the latest version
>> of IF-MIB is from [RFC2863].
>>
>>
>> C) A reference to [RFC2863] needs to be provided in the
>> Normative References.
>>
> [Kalyan>] Will do
>
>> 8) Section 7. VRRP MIB Structure
>>
>>
>> This section should discuss the relationship between the tables.
>>
>> The groups listed are part of the conformance, and while okay
>> to mention these, that really isn't the MIB's structure.
>>
>> NIT: Should have a (3) before "The vrrpAssociatedIpAddrTable"
>>
> [Kalyan>] Will do.
>
>> 9) Section 8. VRRP MIB Table Design
>>
>> The MIB design should focus on indexing the tables such that
>> the table lookups are more advantageous for the device (e.g. router),
>> not the NM app.  The NM app should be able to
>> handle rearranging data for display purposes.
>>
>> I would like to see a discussion of the indexing wrt the device.
>> The warnings from the MIB compiler indicate that there may be a
>> better indexing (table structure) available.  That doesn't mean
>> that you need to resolve the warnings, but I would like to see
>> some text included which shows why the chosen indexes are
>> advantageous for the VRRP router, and the Virtual Routers.
>>
>> IMHO, if an interface goes down,
>> then would be probably more advantageous
>> to know the Virtual router(s) associated with that interface.
>> (and subsequently the IP addresses on those Virtual Routers).
>>
>> Please combine Section 7 and Section 8.
>>
> [Kalyan>] I do not have any strong reasons for the indexing, As I
> mentioned, It was suggested in the earlier meeting and it remained that
> way during all these revisions. I will change this to ifIndex, VrId,
> AddrType
>

So will revisit this in the next email or two.

>> 10) Section 9.
>>
>> Thank you for these examples.
>>
>> 11) Section 10.  Please specify that this is
>> the VRRP MIB Module Definition
>>
>> * DATE and REVISION dates need to be updated.
>>
>> * VrId (Textual Convention)
>>
>> There is already a VrId in RFC2787. While it looks as if the
>> only difference is the DESCRIPTION clause, need to caution against
>> doing redefining this.  I would suggest creating a VRID TC and
>> making the DESCRIPTION simple, for example:  This value uniquely
>> identifies
>> a
>> Virtual Router on a VRRP router.
>>
>> While the remaining text is informative, it is not really necessary.
>> Please also give a REFERENCE clause for this.
>>
> [Kalyan>] Agreed.
>
>> * vrrpNodeVersion
>>
>> This scalar has correctly been deprecated, but I don't see any
>> version object replacement in a table.  Why not?  (There is
>> field in the protocol for this.)
>>
> [Kalyan>] This MIB would support only version 3 of the protocol and hence
> I removed the version from the table.

Okay.

>
>> * vrrpNotificationCntl, vrrpTrapNewMasterCntl, vrrpTrapProtoErrorCntl
>>
>> Could notifications be generated or not, using
>> RFC3413?
>>
> [Kalyan>] Notifications could be filtered/generated using RFC3413. I added
> these to conveniently control notifications within this MIB (Based on
> request from some one on the mailing list) You suggest we getrid of these?
>

Yes, I think that most operators would likely use rfc3413.  Having this in 2
places
may create some confusion...


>> *StorageType objects:
>>
>> If an operator configures the VRRP Router with Virtual Router(s)
>> and IPvX addresses then I suspect that this information would be
>> saved to nonVolatile storage.  Do you agree?
>>
>> In other words, I'm trying to think of a scenario when an operator
>> wouldn't want to save this information (once he/she is satisfied with
>> the configuration) in the sake of simplification, would there be
>> benefit to just state that the configuration should be saved
>> to NV storage?  If so, then
>> the use of StorageType objects in this MIB could be removed and
>> replaced with text in the Table Entry DESCRIPTION.
>>
>> (RFC4181, Section 4.6.4 specifies that StorageType objects can be
>> used OR, the Table Entry DESCRIPTION clause specify what happens
>> to dynamically-created rows after an agent restart.)
>>
> [Kalyan>] I am OK with either way of doing this - Will add the text to
> description and remove the StorageType.
>>

Thanks!

>> * RowStatus Objects
>>
>> There is some discussion about having to have the corresponding
>> Virtual Router in vrrpOperationsState have the value of
>> 'initialize' before changing a value in the Row, and then the
>> RowStatus object is supposed to not be "active(1)", but I don't
>> understand what values are acceptable for modifying a read-creat
>> object in a row?  Also, how does this impact vrrpOperationsState?
>>
>> * NIT: AdminState and OperationsState, could these be renamed to use
>> use AdminStatus and OperStatus?
>>
> [Kalyan>] Will do.
>
>> * vrrpRouterVrIdErrors
>>
>> Please verify that the DESCRIPTION is still valid for this MIB?
>> Should a value be added to the Statistics Table instead (i.e. on
>> a per Virtual Router basis?
>>
> [Kalyan>] vrId itself is invalid, so we probably cant index into the
> table? (as one of the indexes for the table is vrId)? May be I am not
> understanding the problem correctly.
>

Sorry, I meant to ask if this scalar is valid wrt the DESCRIPTION?
Should an Errors object be added to the Statistics Table to record errors
on a per VRRP Router basis?

>> * Please refrain from using the term "trap" in the current
>> objects/notifications.
>> Please use the term "notification" as appropriate for the
>> current objects/notifications.  (Deprecated items are fine to leave the
>> term
>> trap.)
>>
> [Kalyan>] Will do.
>
>> * vrrpNewMasterReason, vrrpTrapProtoErrReason
>>
>> Where are these Enum values from?  Could you add a reference?
>> Also, please explain the values.
>>
> [Kalyan>] These are not defined in the spec. We just came up with possible
> reasons for a VR to transmission to Master state.
>

Okay, please add more description as to how each reason is determined.


>> * Conformance looks fine, but will review again once above MIB
>> Module comments are addressed.  Same with security section.
>>
> [Kalyan>] Thanks again for the detailed review. I probably is a good idea
> to start with a new MIB and address all the issues you raised.
>

Thanks,
  -Joan

>>
>> Thanks,
>>  -Joan
>>
>>
>>
>> _______________________________________________
>> vrrp mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/vrrp
>>
>> Scanned by Check Point Total Security Gateway.
>


Scanned by Check Point Total Security Gateway.
_______________________________________________
vrrp mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/vrrp