Re: [bess] RFC 8584: EVPN DF Election - Originating Router's IP Address of different address family

[Muthu Arul Mozhi Perumal]

Hi Satya,

Please see inline..

On Wed, Feb 10, 2021 at 8:37 AM Satya Mohanty (satyamoh) 
wrote:

> Hi Tulasi and Muthu,
>
>
>
> Yes, the numeric value here refers to the 4byte or 16byte unsigned value
> representation of the IP address field.
>
> Maybe in the 7432-bis this can be stated explicitly.
>

Agree, this is important to clarify in 7432-bis and simple to fix in
implementations that don't compare IPv4 and IPv6 addresses this way for
default DF election algo..


>
>
> We did envision this possibility in RFC 8584 (the HRW hash already takes
> care of the fact that the IP address could be either IPV4 or IPV6).
>
>
>
> However, with respect to deployment do you have a case in mind of this
> “mixed v4/v6” that will necessitate such a tie-break?
>
> I mean one of the MH PEs has an ipv4 originator address and the other a v6
> originator address.
>

Yes, in a dual-stack environment, there could be multihomed PEs running (1)
BGP over IP4 w/ IPv4 NH, (2) BGP over IPv6 w/ IPv4 NH, (3) BGP over IPv6 w/
IPv6 NH, say connected thru' an RR in control plane, and all part of the
same ES group. In this case, the Originating Router's IP address in the ES
route could be an IPv4 or IPv6 address (implementation specific)..

Regards,
Muthu


>
>
> Thanks,
>
> --Satya
>
>
>
>
>
> *From: *BESS  on behalf of TULASI RAM REDDY <
> tulasiramire...@gmail.com>
> *Date: *Friday, February 5, 2021 at 5:17 AM
> *To: *Muthu Arul Mozhi Perumal 
> *Cc: *"bess@ietf.org" 
> *Subject: *Re: [bess] RFC 8584: EVPN DF Election - Originating Router's
> IP Address of different address family
>
>
>
> Thanks Muthu.
>
> Shouldn't numeric value here mean simply the 4byte or 16 byte unsigned
> value representation of the IP address field?
>
> Thinking loudly on how to interpret numeric value  and the limitations
> here. Any comments?
>
>
>
> each PE builds an ordered list of the IP
>
> addresses of all the PE nodes connected to the Ethernet segment
>
> (including itself), in *increasing numeric value*
>
>
>
> Thanks,
>
> Tulasi.
>
> On Fri, Feb 5, 2021 at 3:15 PM Muthu Arul Mozhi Perumal <
> muthu.a...@gmail.com> wrote:
>
> Hi Tulasi,
>
>
>
> I think the problem is, there is no standard way to numerically compare
> IPv4 with IPv6 addresses to form an ordered list. So, all the PEs
> multihomed to an ES may not always arrive at the same DF (or BFD in the
> case of single-active L-LINE service) with the default DF election algo.
> This is problematic (and may cause traffic loops). Hence, the default DF
> election algo works only when all PEs multihomed to an ES have Originating
> Router's IP Address of the same AF..
>
>
>
> Regards,
>
> Muthu
>
>
>
> On Thu, Feb 4, 2021 at 7:58 PM TULASI RAM REDDY 
> wrote:
>
> Hi All,
>
>
>
> As mentioned in the Problem Statement of RFC8584 Sec 1.3, Default DF
> algorithm is expected to have
>
> all multihomed PEs to have Originating IP of the same address family.
>
> Do we see any interop issue if the different address families are
> considered, i.e. ordering in
>
> ascending order based on numerical value in Originating IP here? For IPv4
> read 4 octets as unsigned integer
>
> and IPv6 is considered as 16 octet unsigned integer.
>
>
> 1.3 .  Problem Statement
>
> Default DF election algorithm assumes
>
> that all the PEs who are multihomed to the same ES (and interested in
>
> the DF election by exchanging EVPN routes) use an Originating
>
> Router's IP address [RFC7432 ] of the 
> same family.
>
>
>
> Thanks,
>
> TULASI RAMI REDDY N
>
> ___
> BESS mailing list
> BESS@ietf.org
> https://www.ietf.org/mailman/listinfo/bess
>
>
>
> --
>
> TULASI RAMI REDDY N
>
___
BESS mailing list
BESS@ietf.org
https://www.ietf.org/mailman/listinfo/bess

Re: [bess] RFC 8584: EVPN DF Election - Originating Router's IP Address of different address family

[Satya Mohanty (satyamoh)]

Hi Jorge,

Inline.

From: BESS  on behalf of "Rabadan, Jorge (Nokia - 
US/Mountain View)" 
Date: Friday, February 5, 2021 at 6:05 AM
To: TULASI RAM REDDY , Muthu Arul Mozhi Perumal 

Cc: "bess@ietf.org" 
Subject: Re: [bess] RFC 8584: EVPN DF Election - Originating Router's IP 
Address of different address family

Hi Tulasi,

For the default DF algorithm, I think at that point it will be up to the 
implementation how to resolve the candidate list PEs with originating IPs of 
different family.
 This is should be standardized as otherwise It will result in different 
decisions in different PEs. I suggest a mention in 7432-bis.
I know of some implementations that do what you are saying, but you can’t 
assume all RFC7432/8584 implementations will do that.

Thanks.
Jorge

Thanks,
Satya

From: BESS  on behalf of TULASI RAM REDDY 

Date: Friday, February 5, 2021 at 2:15 PM
To: Muthu Arul Mozhi Perumal 
Cc: bess@ietf.org 
Subject: Re: [bess] RFC 8584: EVPN DF Election - Originating Router's IP 
Address of different address family
Thanks Muthu.
Shouldn't numeric value here mean simply the 4byte or 16 byte unsigned value 
representation of the IP address field?
Thinking loudly on how to interpret numeric value  and the limitations here. 
Any comments?


each PE builds an ordered list of the IP

addresses of all the PE nodes connected to the Ethernet segment

(including itself), in increasing numeric value

Thanks,
Tulasi.
On Fri, Feb 5, 2021 at 3:15 PM Muthu Arul Mozhi Perumal 
mailto:muthu.a...@gmail.com>> wrote:
Hi Tulasi,

I think the problem is, there is no standard way to numerically compare IPv4 
with IPv6 addresses to form an ordered list. So, all the PEs multihomed to an 
ES may not always arrive at the same DF (or BFD in the case of single-active 
L-LINE service) with the default DF election algo. This is problematic (and may 
cause traffic loops). Hence, the default DF election algo works only when all 
PEs multihomed to an ES have Originating Router's IP Address of the same AF..

Regards,
Muthu

On Thu, Feb 4, 2021 at 7:58 PM TULASI RAM REDDY 
mailto:tulasiramire...@gmail.com>> wrote:
Hi All,

As mentioned in the Problem Statement of RFC8584 Sec 1.3, Default DF algorithm 
is expected to have
all multihomed PEs to have Originating IP of the same address family.
Do we see any interop issue if the different address families are considered, 
i.e. ordering in
ascending order based on numerical value in Originating IP here? For IPv4 read 
4 octets as unsigned integer
and IPv6 is considered as 16 octet unsigned integer.

1.3.  Problem Statement

Default DF election algorithm assumes

that all the PEs who are multihomed to the same ES (and interested in

the DF election by exchanging EVPN routes) use an Originating

Router's IP address [RFC7432] of the same 
family.

Thanks,
TULASI RAMI REDDY N
___
BESS mailing list
BESS@ietf.org
https://www.ietf.org/mailman/listinfo/bess


--
TULASI RAMI REDDY N
___
BESS mailing list
BESS@ietf.org
https://www.ietf.org/mailman/listinfo/bess

Re: [bess] RFC 8584: EVPN DF Election - Originating Router's IP Address of different address family

[Satya Mohanty (satyamoh)]

Hi Tulasi and Muthu,

Yes, the numeric value here refers to the 4byte or 16byte unsigned value 
representation of the IP address field.
Maybe in the 7432-bis this can be stated explicitly.

We did envision this possibility in RFC 8584 (the HRW hash already takes care 
of the fact that the IP address could be either IPV4 or IPV6).

However, with respect to deployment do you have a case in mind of this “mixed 
v4/v6” that will necessitate such a tie-break?
I mean one of the MH PEs has an ipv4 originator address and the other a v6 
originator address.

Thanks,
--Satya


From: BESS  on behalf of TULASI RAM REDDY 

Date: Friday, February 5, 2021 at 5:17 AM
To: Muthu Arul Mozhi Perumal 
Cc: "bess@ietf.org" 
Subject: Re: [bess] RFC 8584: EVPN DF Election - Originating Router's IP 
Address of different address family

Thanks Muthu.
Shouldn't numeric value here mean simply the 4byte or 16 byte unsigned value 
representation of the IP address field?
Thinking loudly on how to interpret numeric value  and the limitations here. 
Any comments?


each PE builds an ordered list of the IP

addresses of all the PE nodes connected to the Ethernet segment

(including itself), in increasing numeric value

Thanks,
Tulasi.
On Fri, Feb 5, 2021 at 3:15 PM Muthu Arul Mozhi Perumal 
mailto:muthu.a...@gmail.com>> wrote:
Hi Tulasi,

I think the problem is, there is no standard way to numerically compare IPv4 
with IPv6 addresses to form an ordered list. So, all the PEs multihomed to an 
ES may not always arrive at the same DF (or BFD in the case of single-active 
L-LINE service) with the default DF election algo. This is problematic (and may 
cause traffic loops). Hence, the default DF election algo works only when all 
PEs multihomed to an ES have Originating Router's IP Address of the same AF..

Regards,
Muthu

On Thu, Feb 4, 2021 at 7:58 PM TULASI RAM REDDY 
mailto:tulasiramire...@gmail.com>> wrote:
Hi All,

As mentioned in the Problem Statement of RFC8584 Sec 1.3, Default DF algorithm 
is expected to have
all multihomed PEs to have Originating IP of the same address family.
Do we see any interop issue if the different address families are considered, 
i.e. ordering in
ascending order based on numerical value in Originating IP here? For IPv4 read 
4 octets as unsigned integer
and IPv6 is considered as 16 octet unsigned integer.

1.3.  Problem Statement

Default DF election algorithm assumes

that all the PEs who are multihomed to the same ES (and interested in

the DF election by exchanging EVPN routes) use an Originating

Router's IP address [RFC7432] of the same 
family.

Thanks,
TULASI RAMI REDDY N
___
BESS mailing list
BESS@ietf.org
https://www.ietf.org/mailman/listinfo/bess


--
TULASI RAMI REDDY N
___
BESS mailing list
BESS@ietf.org
https://www.ietf.org/mailman/listinfo/bess