Re: [spring] packet captures for draft-ietf-spring-srv6-network-programming-06?

[Andrew Alston <[e-mail]>]

Echoing what Mark said in later emails.  A deployment draft was published – it made claims about peoples production deployments – and it is my belief that any draft that is put forward into the IETF is open to testing and verification.  Furthermore – I would strongly argue there is a reason that the IETF calls for running code, running code lets us test the premise behind the drafts – and if the running code presents – and further pushed through deployment drafts – is not in line with the drafts themselves – we need to understand why not.

To find inter-operable approaches – which is kinda the point of standards I believe – it is very critical that we understand what is actually being put out there as being deployed – vs what is in the drafts – and critical to understand if we are miss understanding something in the draft that is causing such wide disparities – anything else will lead to a huge amount of inter-op issues later down the line.

Hence – I think an analysis of what is being claimed by the deployment draft is indeed very relevant.

Thanks

Andrew


From: Robert Raszuk <[e-mail]>
Date: Tuesday, 17 December 2019 at 13:12
To: Andrew Alston <[e-mail]>
Cc: Alexandre Petrescu <[e-mail]>, SPRING WG email list <[e-mail]>
Subject: Re: [spring] packet captures for draft-ietf-spring-srv6-network-programming-06?

Hi Andrew,

My personal opinion is that with below you are now going way outside of what should be discussed on IETF mailing lists. Hope SPRING charis will address it. IETF is not the right forum for any vendor implementation discussion regardless if this is Cisco, Juniper, Arrcus, Nokia etc .... I recommend you move it to -nsp lists.

If standards or drafts are not clear you are welcome to ask questions on those. Any implementation is a private choice of given vendor and in no way should influence WG decision in regards of the choices we make in protocol design.

If you think that some implementations violate standards or even WG drafts you are more then welcome to propose specific questions to the implementation reports which chairs would be normally more than happy to include in the process and ask or even enforce all vendors to fill the blanks.

Regards,
Robert.


On Tue, Dec 17, 2019 at 6:58 AM Andrew Alston <[e-mail]<mailto:[e-mail]>> wrote:
Alex,

Will try and get you some captures off the devices I’ve been testing on – in order to make sure I understood this draft properly, and in light of the deployment status draft, I decided to play a lot more deeply and setup a bit of a lab.  I’m still doing tests and soon as I have some other bits completed will send through the packet captures from those against (Since the XR boxes that I have to test on seem to have absolutely no ability to setup traffic steering with SRv6 (and I actually have requested details of how to configure this in the past but gotten no response), I’m just finishing the code to inject packets from outside with a sid stack to test this.  I also acknowledge that I’m running tests against code that is implementing a draft that seems far from final – and so shouldn’t have that many expectations.

That being said, In light of the deployment draft – I do have some concerns that there is a draft that specifies that people have put this stuff into production – yet the implementation in current shipping code seems to be *way* off the draft and contrary to things we have been told in the working group.

Some of the more interesting finds so far:


  *   In Montreal – I questioned the growth in the IGP tables – since I would have to use a separate locator on each router – I was explicitly told this wasn’t necessary and could use the loopbacks – not so in current code – use of the loopback marks the locator as down.


  *   Locator size is not configurable as anything other than a /64



  *   XR 7.0.1 claims a maximum number of SID’s at 8000 – I’m still unclear if this limitation in the code is based on locally configured SID’s or received SID’s – and will run some tests on this in the coming day or two to verify



  *   There seems to be a limit on a single locator per box – I’m still trying to figure out what impact this will have in a multi-area or multi-level IGP deployment scenario.



  *   By default when configuring a locator – the device configures a separate End.X (PSP) for each interface – now – this is where things get interesting.  If I am reading the NP text correctly, End.X (PSP) should be locator:0006::  - However, in the shipping code, that is not the case at all – as per the below:



RP/0/RP0/CPU0:SRV6-R2#show segment-routing srv6 locator R2 sid Sun Dec 15 04:56:10.913 UTC

SID                         Behavior     Context                           Owner               State  RW

--------------------------  -----------  ------------------------------    ------------------  -----  --

2001:db8:ee:2:1::           End (PSP)    'default':1                       sidmgr              InUse  Y

2001:db8:ee:2:11::          End.OP       'default'                         sidmgr              InUse  Y

2001:db8:ee:2:40::          End.X (PSP)  [Gi0/0/0/0, Link-Local]           isis-64             InUse  Y

2001:db8:ee:2:41::          End.X (PSP)  [Gi0/0/0/1, Link-Local]           isis-64             InUse  Y

2001:db8:ee:2:42::          End.X (PSP)  [Gi0/0/0/3, Link-Local]           isis-64             InUse  Y



So from my perspective – I have to wonder about the production deployments – because particularly on this last point – if people have been putting this stuff in production, and the implementation is so different from the text, its going to create some rather interesting breakage going forward if my reading of the text is correct.



Anyway – will send some packet captures hopefully in the next 48 hours once I’ve got a more complete set of captures from my lab setup.



Thanks



Andrew



From: spring <[e-mail]<mailto:[e-mail]>> On Behalf Of Alexandre Petrescu
Sent: Monday, 16 December 2019 17:34
To: SPRING WG email list <[e-mail]<mailto:[e-mail]>>
Subject: [spring] packet captures for draft-ietf-spring-srv6-network-programming-06?

Hi, SPRINGers,

My comments on SRv6 relate to a worry about modifying packets in transit.

In order to better explain myself, or maybe to remove the worry
altogether, I would like to ask for packet dumps of SRv6.

By looking at the packet contents that go into the network it is much
easier to clarify and to avoid misunderstandings.

Alex

_______________________________________________
spring mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/spring<https://www.ietf.org/mailman/listinfo/spring>
_______________________________________________
spring mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/spring<https://www.ietf.org/mailman/listinfo/spring>