<div dir="ltr"><div><br></div>While it's great to have fast Producer -> RIB -> FIB propagation and installation let me ask why does it really matter in practice ? <div><br></div><div>Wouldn't it be much better to pre-install backup paths in FIB ahead of any failures ? For basic BGP PIC Core and Edge could be used, for more complex topologies TI-LFA seems to be decent candidate. Especially on new platforms where FIB sizes these days should have sufficient capacity and when some of them even support native hierarchy ?</div><div><br></div><div>It always makes me wonder why so many folks still count on protocol convergence and router "convergence" upon any bad event instead of just applying few more lines of config and let protocols or for that matter platforms converge peacefully ? </div><div><br></div><div>Kind regards,</div><div>R.</div></div><br><div class="gmail_quote"><div dir="ltr">On Mon, Oct 8, 2018 at 6:14 AM Robert Hass <<a href="mailto:robhass@gmail.com">robhass@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div>I've got 2 full DFZ feeds from my other routers.</div><div>Config - basic - just 2 BGP session and proper rules to not re-announce between both eBGP neighbors.<br></div><div> After "clear ip bgp neighbor *" full convergence FIB+RIB is above 3 minutes.</div><div><br></div><div>Used telemetry + CLI to look how much of prefixes has been installed in FIB to determine above time.</div><div><br></div><div>If I went from 2 neighbors to 1 neighbor result was similar.</div><div><br></div></div><br><div class="gmail_quote"><div dir="ltr">On Sun, Oct 7, 2018 at 5:07 PM Patrik Olsson <<a href="mailto:darkiesan@me.com" target="_blank">darkiesan@me.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">> I just tested RIB convergence time and it's far for being superb... Cisco and JNPR is much faster here.<br>
> Running latest 4.20 and 4.21 releases.<br>
> <br>
<br>
What happens when you test RIB and FIB convergence as two combined events?<br>
<br>
How does the lab setup look like? What feeds the 7280R with the full feed(s)?<br>
What is the conf of the elements?<br>
<br>
Patrik<br>
<br>
</blockquote></div>
-- <br>
arista-nsp mailing list<br>
<a href="mailto:arista-nsp@puck.nether.net" target="_blank">arista-nsp@puck.nether.net</a><br>
<a href="https://puck.nether.net/mailman/listinfo/arista-nsp" rel="noreferrer" target="_blank">https://puck.nether.net/mailman/listinfo/arista-nsp</a><br>
</blockquote></div>