>>> Are you saying you need to know about import complete about per-route as 
>>> well ?
No, I think that would be too much.  We just need this indication to at the VRF 
level. Just an indication of to/from table name would be good enough.

>>> Peer configured down
Yes 10-15 minutes seems reasonable, but I think we need to provide full context 
to the station about this peer for any future state change. For example, 
another indication if this peer is removed completely by the user. If the peer 
comes up then anyway peer UP would be sent.

>>> BMP health
Yes, there needs to some other high priority thread/or other channel to report 
CPU constraints which is causing slow BMP feed generation. That is probably 
implementation detail. Other scenario is when BMP may consume lot of memory 
because it may build up lot of state internally due to slow feed generation. 
This can be used for network operator to take corrective action.

Thanks
Mukul
From: Narasimha Prasad S N (snprasad) <[email protected]>
Date: Thursday, November 6, 2025 at 7:13 PM
To: Srivastava, Mukul <[email protected]>, 
[email protected] <[email protected]>
Subject: RE: Some comments from IETF-124

Hi Mukul

Thank you for reaching out and sharing your comments, sorry you couldn’t ask 
your questions during the preso..


  1.  Route Import Complete use-case came from Jeff Haas, so happy to extend it 
with filters based on both your needs.
     *   The example we shared, indicates if the import activity on the router 
is complete, it is a coarse event (no filters).
     *   From a filters pov, we were thinking may be VRF/RD could be added to 
indicate amongst which tables/VRFs is the import complete (so extension may be 
required to indicate From/To etc).  We would prefer to keep this at the most to 
a VRF/table level.
     *   Are you saying you need to know about import complete about per-route 
as well ? That would make this message per-route then. Scalability would be an 
issue if we start sending this per-route I think.  But happy to hear more on 
the use-case from you and see if we can accommodate it here or if not use the 
REL message draft.


  1.  Peer configured down – we have suggested the time for down needs to be 
configurable on the router (in Ops considerations), something long enough to 
avoid false triggers like you bringup. I have heard from Luuk something like 
10-15 minutes down is what they had in mind & we agree that would be a long 
enough time. What do you think ?



  1.  BMP Health – this is an interesting one. Could you please share more on 
what you had in mind for the different health metrics please.



     *   When you say lack of CPU, are you suggesting Router send a Health 
event, that BMP couldn’t send messages because of lack of CPU or other reason 
code on why it is stuck ? Since this is a BMP message, for it not to be stuck 
because of lack of CPU or other things, presume some out of band/ higher 
priority thread would be used to send the health event ?

Thanks.
Prasad







Cisco Confidential
From: Srivastava, Mukul <[email protected]>
Sent: 07 November 2025 04:38
To: [email protected]
Subject: Some comments from IETF-124

Hello authors

I had below comments during the presentation which I wasn’t able ask due to 
time constraints.


  1.  Route Import Complete: Can you clarify a use case where this event would 
be useful for a network operator who is monitoring a network. If we want to 
make this event useful then I would assume it needs some more details like -

     *   From which RIB route has been imported.

     *   To which RIB route has been imported.

     *   Etc.

  1.
      2.  Peer Configured is Down:  How long a router waits before sending this 
event for a configured BGP peer? The BGP peer may take time to come up and it 
may not be very useful to send this event to station too early.

The other event that I think would be useful to add is “BMP health” event.
This event can be used by router to indicate the BMP station about its BMP 
operation health, for example,  BMP process on the router might be stuck or not 
getting enough CPU to generate BMP feed.

Thanks
Mukul

_______________________________________________
GROW mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to