Passing default route through BGP?

13 Apr 2015

Hey Mike,

Sorry, the specific issue I'm having is around this paragraph:

>If the default route is learned from your upstream, then I would have
>expected that you need that set to allow it to be re-advertised to your igp
>peers - are you saying that it is /not/ the case?  That
>originate-default=if-installed will re-advertise the default route
>regardless?

The issue is that say you have three routers, each distributing a default
route to each other (each has their own upstream, but you want all routers
to have a path to the internet if their upstream goes down).  As soon as
you set originate-default=if-installed, that peer does not *receive* a
default route on that session.

say you have rtr01, rtr02, rtr03.
rtr01 has IBGP_RTR02, and IBGP_RTR03
rtr02 has IBGP_RTR01, and IBGP_RTR03
rtr03 has IBGP_RTR01 and IBGP_RTR02

if I set both peers on rtr01 to originate-default=if-installed, then rtr02
and rtr03 receive default routes from their upstreams AND rtr01

Then if I go to rtr02 and set originate-default=if-installed on IBGP_RTR01
peer, rtr02 stops receiving a default route from rtr01, and if it's
upstream goes down, it now has no default route.

I might lab it up in GNS3 tonight and see if cisco gives the same result -
I swear it doesn't, but at this point I'm just darned confused, so maybe it
does :-p

Yeah, i've noticed the CPU pegging issue - that's why I was only taking
default + one-hop-in-AS-path from each provider - a full feed takes ~3 mins
to fully process a session up or down on a CCR1009, which leaves a lot of
downtime.  Been crossing my fingers that V7 comes out soon, because I
*really* want to use RouterOS ;)  (plus four CCR1009's would take up less
space than four 7206's, and look a whole lot sexier)

Cheers,

Damien

On 15 April 2015 at 11:18, Mike Everest <mike@duxtel.com> wrote:

> Hi Damien,
>
> I'm not sure that I fully understand the behaviour that you are reporting
> :-}
>
> redistribute-other-bgp is intended to allow the router to advertise to its
> peers route prefixes that it has learned from other peers - which instance
> is involved instance itself shouldn't matter (don't think so, but haven't
> tested it! ;)
>
> If the default route is learned from your upstream, then I would have
> expected that you need that set to allow it to be re-advertised to your igp
> peers - are you saying that it is /not/ the case?  That
> originate-default=if-installed will re-advertise the default route
> regardless?
>
> If you are saying that the learned route is re-advertised with the nexthop
> of the upstream router, then you need to force the nexthop to 'self' - look
> for 'nexthop-choice=force-self' attribute under bgp peer settings.
>
> But I get the feeling that I'm missing the important point of your thread -
> if so, sorry about that! :-}
>
> Regarding alternative platforms, I'm with you on that - routerOS is just
> too
> good to be easily overcome by lower price - but that's my own opinion, of
> course, and you'd expect someone who runs the largest Australia/pacific
> distributor of a product to have that opinion! :-D
>
> About RAM size for BGP tables - the answer is 'a lot' ;)  I don't have any
> routers myself that contain even one full table, so I can't tell you
> exactly
> what a full table costs in RAM, but I'm sure that there are others on this
> list that can tell you ;)
>
> What I do know is that when it comes to multiple BGP peers with full global
> tables, CPU cost is a lot more important than RAM!  The problem is that
> current routerOS (v6) is still limited to one thread for routing table
> update task, so you end up with one CPU core always pegged, and routing
> table update lagging.  In one case that I am aware of, it caused sufficient
> grief that the network operator had no choice than to take out the CCRs and
> replace with other routers.  Yes, they cost a lot more (A LOT MORE! ;) but
> CCR simply couldn't keep up with that demand.
>
> I don't know for sure what the actual demand was for that particular case,
> but again, someone else on the list might be able to offer some real-world
> data on that!
>
> Cheers,  Mike.
>
> > -----Original Message-----
> > From: Public [mailto:public-bounces@talk.mikrotik.com.au] On Behalf Of
> > Damien Gardner Jnr
> > Sent: Tuesday, 14 April 2015 10:32 PM
> > To: MikroTik Australia Public List
> > Subject: Re: [MT-AU Public] Passing default route through BGP?
> >
> > Thanks Mike,
> >
> > If I set it to originate a default, then it does originate it.  After
> some
> playing, if
> > I set it to originate-default=if-installed, then it is
> > *passing* the default route it has learnt.  However once a session is
> > originating default, it will not receive default from that same peer.
> >
> > I just set it up again, rtr01, originating-default if-installed to rtr03,
> and rtr03
> > sees the defaults from its upstream and rtr01 (and the static
> > route)l:
> >
> > [admin@rtr03] > /ip route print detail where dst-address=0.0.0.0/0
> > Flags: X - disabled, A - active, D - dynamic, C - connect, S - static, r
> -
> rip, b -
> > bgp, o - ospf, m - mme, B - blackhole, U - unreachable, P - prohibit
> >  0 ADb  dst-address=0.0.0.0/0 gateway=223.25.113.57
> > gateway-status=223.25.113.57 reachable via  vlan308 distance=20
> >         scope=40 target-scope=10 bgp-as-path="55707" bgp-local-pref=50
> bgp-
> > origin=igp bgp-communities=54320:55707
> >         received-from=BGP_IPv4_AS55707
> >  1   S  dst-address=0.0.0.0/0 gateway=223.25.113.57
> > gateway-status=223.25.113.57 reachable via  vlan308 distance=200
> >         scope=30 target-scope=10
> >  2  Db  dst-address=0.0.0.0/0 gateway=103.235.52.65
> > gateway-status=103.235.52.65 reachable via  vlan99 distance=200
> >         scope=40 target-scope=30 bgp-as-path="9482,38285,1221"
> > bgp-local-pref=50 bgp-origin=igp
> >
> > bgp-
> > communities=9482:14201,9482:19201,9482:65500,38285:4,38285:20,38285:22,
> > 38285:1221,54320:9482
> >         received-from=IBGP_IPv4_RTR01
> >
> >
> > Then I set originate-default if-installed on rtr03, and it's now only
> seeing the
> > default route from its upstream (and the static route):
> >
> > [admin@rtr03] > /ip route print detail where dst-address=0.0.0.0/0
> > Flags: X - disabled, A - active, D - dynamic, C - connect, S - static, r
> -
> rip, b -
> > bgp, o - ospf, m - mme, B - blackhole, U - unreachable, P - prohibit
> >  0 ADb  dst-address=0.0.0.0/0 gateway=223.25.113.57
> > gateway-status=223.25.113.57 reachable via  vlan308 distance=20
> >         scope=40 target-scope=10 bgp-as-path="55707" bgp-local-pref=50
> bgp-
> > origin=igp bgp-communities=54320:55707
> >         received-from=BGP_IPv4_AS55707
> >  1   S  dst-address=0.0.0.0/0 gateway=223.25.113.57
> > gateway-status=223.25.113.57 reachable via  vlan308 distance=200
> >         scope=30 target-scope=10
> >
> >
> > I don't have redistribute-other-bgp set to yes, as the documentation
> seems
> > to suggest that it is only learnt to share routes between BGP instances -
> and
> > I'm only running one instance on each router?  I did try turning it on
> tonight,
> > and it made no difference though..
> >
> > The only filter I have is an accept filter on the _in filter for the iBGP
> session,
> > to only accept 0.0.0.0/0 if the 54320:9482 etc community is set, as I
> was
> > randomly getting a default route via IBGP which didn't have bgp info on
> it
> > (bgp-origin being 'incomplete') - I'm assuming that was when the other
> peer
> > was sending it's static default route as well as it's bgp learnt route?
> >
> > Although a slightly related question which may render this all not overly
> > needed (though would still be nice to have it working) - how much ram do
> I
> > need to store four full BGP tables (IPv4 + IPv6) per router?  I figured
> it
> would
> > mean needing 2gb of ram, since one full feed of both requires 512mb, but
> > reading the Mikrotik BGP FAQ, it looks like maybe I could get away with
> 1gb
> > of ram?  Though that's probably still leaving me needing a CCR as each
> edge
> > router, as I don't think there's anything 'smaller' with that much
> > ram? Even getting 512mb seems difficult.      (Edgerouter Lite's are
> > starting to look like a good cost point ($150 for 512mb and 3 x gig
> ports,
> and
> > they actually will route gigabit), but I *really* like the CLI on
> routerOS, the
> > REST API is awesome, and WinBox for quick changes is
> > fantastic..)
> >
> > Thanks,
> >
> > Damien
> >
> >
> >
> > On 14 April 2015 at 09:09, Mike Everest <mike@duxtel.com> wrote:
> >
> > > Hi Damien,
> > >
> > > Are you sure it's not being filtered out somewhere?  RouterOS
> > > definitely advertises default route when configured to do so (i.e.
> > > default-originate=if-installed) Of course you would also need to
> > > redistribute routes learned from other BGP too (i.e. instance has
> > > redistribute-other-bgp=yes set), but judging from your comments thus
> > > far, I guess you already know that :-}
> > >
> > > Does default route show up in "rout bgp ad pr"?
> > >
> > > Cheers!
> > >
> > > Mike.
> > >
> > > > -----Original Message-----
> > > > From: Public [mailto:public-bounces@talk.mikrotik.com.au] On Behalf
> > > > Of Damien Gardner Jnr
> > > > Sent: Monday, 13 April 2015 8:55 PM
> > > > To: MikroTik Australia Public List
> > > > Subject: [MT-AU Public] Passing default route through BGP?
> > > >
> > > > Hi Guys,
> > > >
> > > > Our primary upstream has had a couple of small periods of complete
> > > > downtime the last couple of nights, which has given me the
> > > > opportunity
> > > (lol)
> > > > to test our failover with our other two upstreams.
> > > >
> > > > Found an issue in that it looks like RouterOS doesn't pass default
> > > > route through BGP?
> > > >
> > > > I currently have one router setup per upstream, receives a default
> > > > route
> > > plus
> > > > full feed (but we filter to only directly-connected (one AS hop)
> > > > prefixes
> > > for
> > > > each upstream)
> > > >
> > > > I then run IBGP between all edge routers.  And also run OSPF between
> > > > the four edge routers, and the core router, to send a default route
> > > > to the
> > > core.
> > > >
> > > > My thinking was that OSPF will redistribute a default route on any
> > > > edge router with an active default route (i.e. BGP is up and/or
> > > > gateway IP is
> > > > pingable)  Which works ok, IF the upstream gateway goes down.  If I
> > > remove
> > > > the hardcoded default routes, then it works as expected, but leaves
> > > > those routers uncontactable.
> > > >
> > > > I had also been expecting that the default route from each upstream
> > > > was being passed on via iBGP to the other routers - to the point
> > > > that I had
> > > been
> > > > setting localpref on each one based on their cost and available
> > > bandwidth.
> > > >
> > > > On checking tonight, it seems that the 0.0.0.0/0 routes are NOT
> > > > being propagated via IBGP?
> > > >
> > > > I can set default-originate=if-installed, which sends a default
> > > > route
> > > from
> > > one
> > > > peer to another, but it is not the one being received on eBGP from
> > > > the usptream, as the localpref is wrong.  And if I set
> > > default-originate=if-installed
> > > > on the other peer, that peer stops receiving that default route.
> > > >
> > > > Has anyone tried this?  It works fine on quagga and cisco, but
> > > > doesn't
> > > appear
> > > > to be working on mikrotik (which if it's correct, makes me glad I
> > > > found
> > > it
> > > > before I'd forked out on another three hardware routers to go
> > > > alongside
> > > my
> > > > CCR when it arrives back)
> > > >
> > > > Thanks,
> > > >
> > > > DG
> > > >
> > > >
> > > > --
> > > >
> > > > Damien Gardner Jnr
> > > > VK2TDG. Dip EE. GradIEAust
> > > > rendrag@rendrag.net -  http://www.rendrag.net/
> > > > --
> > > > We rode on the winds of the rising storm,  We ran to the sounds of
> > > thunder.
> > > > We danced among the lightning bolts,  and tore the world asunder
> > > > _______________________________________________
> > > > Public mailing list
> > > > Public@talk.mikrotik.com.au
> > > > http://talk.mikrotik.com.au/mailman/listinfo/public_talk.mikrotik.co
> > > > m.au
> > >
> > >
> > > _______________________________________________
> > > Public mailing list
> > > Public@talk.mikrotik.com.au
> > > http://talk.mikrotik.com.au/mailman/listinfo/public_talk.mikrotik.com.
> > > au
> > >
> >
> >
> >
> > --
> >
> > Damien Gardner Jnr
> > VK2TDG. Dip EE. GradIEAust
> > rendrag@rendrag.net -  http://www.rendrag.net/
> > --
> > We rode on the winds of the rising storm,  We ran to the sounds of
> thunder.
> > We danced among the lightning bolts,
> >  and tore the world asunder
> > _______________________________________________
> > Public mailing list
> > Public@talk.mikrotik.com.au
> > http://talk.mikrotik.com.au/mailman/listinfo/public_talk.mikrotik.com.au
>
>
> _______________________________________________
> Public mailing list
> Public@talk.mikrotik.com.au
> http://talk.mikrotik.com.au/mailman/listinfo/public_talk.mikrotik.com.au
>

-- 

Damien Gardner Jnr
VK2TDG. Dip EE. GradIEAust
rendrag@rendrag.net -  http://www.rendrag.net/
--
We rode on the winds of the rising storm,
 We ran to the sounds of thunder.
We danced among the lightning bolts,
 and tore the world asunder

Damien Gardner Jnr

Mike Everest

Damien Gardner Jnr

Mike Everest

Damien Gardner Jnr

tags

participants (2)