Re: BFD WG adoption for draft-haas-bfd-large-packets

["Albert Fu (BLOOMBERG/ 120 PARK)" <[e-mail]>]

Hi Acee,

Please see comments in-line. 

Thanks,

Albert

From: [e-mail] At: 10/23/18 13:02:49To:  Albert Fu (BLOOMBERG/ 120 PARK ) ,  [e-mail],  [e-mail]
Subject: Re: BFD WG adoption for draft-haas-bfd-large-packets

      

Hi Albert,  
  

From: "Albert Fu (BLOOMBERG/ 120 PARK)" <[e-mail]>
Reply-To: Albert Fu <[e-mail]>
Date: Tuesday, October 23, 2018 at 12:45 PM
To: "[e-mail]" <[e-mail]>, "Les Ginsberg (ginsberg)" <[e-mail]>, Acee Lindem <[e-mail]>
Subject: Re: BFD WG adoption for draft-haas-bfd-large-packets 

  

Hi Acee,  

  

You are right in that this issue does not happen frequently, but when it does, it is time consuming to troubleshoot and causes unnecessary network downtime  to some applications (e.g. between two end hosts, some applications worked fine, but others would intermittently fail when they tried to send large size packets over the failing ECMP path). 
  
So you’re saying there is a problem where the data plane interfaces do not support the configured MTU due to a SW bug? I hope these are not our routers 😉


AF> There's no bug.  


1) The issue we have seen is with the Telco network. The router can happily transmit and receive up to configured interface MTU, but the Telco circuit fails to support it. One example is when Telco uses L2VPN to deliver the P2P service to us, but due to some faults, traffic was re-routed to a mis-configured path that did not support our MTU size (e.g. MTU on Telco PE router was not increased to account for MPLS headers for the L2VPN service).


2) AFAIK, the OSPF MTU detection is based on checking MTU value in the DBD packet, The actual OSPF packet size may be smaller and may not detect data plane issue in Telco network during OSPF session establishment.


 

  

I believe the OSPF MTU detection is a control plane mechanism to check config, and may not necessary detect a data plane MTU issue (since OSPF does not support  padding). Also, most of our issues occurred after routing adjacency had been established, and without any network alarms. 
  
Right. However, if the interface is flapped when the MTU changes, OSPF would detect dynamic MTU changes (e.g., configuration), that the control plane is aware of. 


AF> We have encountered the MTU issue without any interface flaps on our routers (no config change on our routers). The MTU issue occurred within the Telco network. Note also some Telco providers that provide WAN circuit spanning several countries may make use of smaller local providers to provide the last mile.  We have seen issues with the smaller providers.


 
  
Thanks, 
Acee  

  

Thanks 

Albert 

  

From: [e-mail] At: 10/23/18 12:30:55 
To: Albert Fu (BLOOMBERG/ 120 PARK ) , [e-mail], [e-mail]
Subject: Re: BFD WG adoption for draft-haas-bfd-large-packets 


Hi Albert, Les,  
  
I tend to agree with Les that BFD doesn’t seem like the right protocol for this. Note that if you use OSPF as your IGP and flap the interface when the MTU changes, you’ll detect MTU mismatches  immediately due to OSPF’s DB exchange MTU negotiation. Granted, control plane detection won’t detect data plane bugs resulting in MTU fluctuations but I don’t see this as a frequent event. 
  
Thanks, 
Acee 
  
  

From: Rtg-bfd <[e-mail]> on behalf of "Albert Fu (BLOOMBERG/ 120 PARK)" <[e-mail]>
Reply-To: Albert Fu <[e-mail]>
Date: Tuesday, October 23, 2018 at 11:44 AM
To: "[e-mail]" <[e-mail]>, "Les Ginsberg (ginsberg)" <[e-mail]>
Subject: RE: BFD WG adoption for draft-haas-bfd-large-packets 

  

Hi Les,   

  

Given that it takes relative lengthy time to troubleshoot the MTU issue, and the associated impact on customer traffic, it is important to have a reliable and fast mechanism to detect the  issue.  

  

I believe BFD, especially for single hop control-plane independent situation (btw, this covers majority of our BFD use case), is indeed an ideal and reliable solution for this purpose. It  is also closely tied with the routing protocols, and enable traffic to be diverted very quickly.  

  

The choice of BFD timer is also one of the design tradeoffs - low BFD detection timer will cause more network churns. We do not need extremely aggressive BFD timer to achieve fast convergence.  For example, with protection, we can achieve end to end sub-second convergence by using relatively high BFD interval of 150ms.  

  

In the case where the path will be used for a variety of encapsulations (e.g. Pure IP and L3VPN traffic), we would set the BFD padding to cater for the largest possible payload. So, in our  case, our link needs to carry a mix of pure IP (1500 max payload) and MPLS traffic (1500 + 3 headers), we would set the padding so that the total padded BFD packet size is 1512 bytes.  

  

As you rightly pointed out, ISIS routing protocol does support hello padding, but since this is a control plane process, we can not use aggressive timer. The lowest hello interval the can  be configured is 1s, so with default multiplier of 3, the best we can achieve is 3s detection time. 

  

What we would like is a simple mechanism to validate that a link can indeed carry the expected max payload size before we put it into production. If an issue occurs where this is no longer  the case (e.g. due to outages or re-routing within the Telco circuit), we would like a reliable mechanism to detect this, and also divert traffic around the link quickly. I feel BFD is a good method for this purpose. 

  

Thanks 

Albert   

  

From: [e-mail] At: 10/23/18 10:45:02 
To: Albert Fu (BLOOMBERG/ 120 PARK ) , [e-mail]
Subject: RE: BFD WG adoption for draft-haas-bfd-large-packets 


Albert – 
  
Please understand that I fully agree with the importance of being able to detect/report MTU issues. In my own experience this can be a difficult problem to diagnose. You do not have  to convince me that some improvement in detection/reporting is needed. The question really is whether using BFD is the best option. 
  
Could you respond to my original questions – particularly why sub-second detection of this issue is a requirement? 
  
For your convenience: 
  
<snip> 
It has been stated that there is a need for sub-second detection of this condition – but I really question that requirement.  
What I would expect is that MTU changes only occur as a result of some maintenance operation (configuration change, link addition/bringup, insertion of a new box in the physical  path etc.). The idea of using a mechanism which is specifically tailored for sub-second detection to monitor something that is only going to change occasionally seems inappropriate. It makes me think that other mechanisms (some form of OAM, enhancements to  routing protocols to do what IS-IS already does J) could be more appropriate  and would still meet the operational requirements. 
  
I have listened to the Montreal recording – and I know there was discussion related to these issues (not sending padded packets all the time, use of BFD echo, etc.) – but I would  be interested in more discussion of the need for sub-second detection. 
  
Also, given that a path might be used with a variety of encapsulations, how do you see such a mechanism being used when multiple BFD clients share the same BFD session and their  MTU constraints are different? 
<end snip> 
  
Thanx. 
  
   Les