Juniper BGP Convergence Time

Saku Ytti saku at ytti.fi
Thu May 31 10:34:48 UTC 2018


Hey Adam,


On 30 May 2018 at 22:49, Adam Kajtar <akajtar at wadsworthcity.org> wrote:

> *Receiving Full Routes*
>
> Convergence time was 180 seconds. The routing table updated and showed the
> correct path in under a minute but the forwarding table took 180 seconds
> for most the routes to update.

How as this verified? Juniper is known to converge on software much
faster than on hardware. You'd need to at least verify:

- show krt queue  => no updates queued
- show system processes extensive |match rpd   => state is 'kqread' not 'RUN'


180seconds seems unlikely for MX80/MX104, may in scenario where box
doesn't have anything else to do, no where to sends routes out to, no
inferior routes to replace with new superior routes etc. Like if you
just have empty box not connected anywhere receiving eBGP from on
peer, maybe. In actual box doing actual work, unlikely.
We see 20-30min convergence times on large boxes at initial
convergency, mostly due to full-mesh being chatty and causing lot of
useless state changes as new data keeps coming in.


-- 
  ++ytti



More information about the NANOG mailing list