Mailing List Archive: show ip[v6] bgp ?

show ip[v6] bgp ?

Nov 11, 2013, 6:54 AM

Post #1 of 4 (1585 views)

In troubleshooting what I believe is a bug in 5.4.0d in which an IPv6 route
claims to be advertised to peers but is not received by those peers, I think
I've found another issue that's more of a feature change than bug. But I'm
curious, in the output of show ip bgp route detail, there are two sections near
the end of the output...

"Route is advertised to X peers:" is pretty clear in what it means.

"Route is to be sent to Y peers:" I can't find any documentation saying what
this means...and I'm curious, what does this mean?

The above mentioned bug was that an ipv6 route for which there was a network
statement and static route to null0, the show ipv6 bgp route detail output said
that the route was being advertised to a subset of the peers the config should
have caused it to be sent to. Those peers were not actually receiving the
route. Removing and re-adding the static route to null0 caused the route to be
advertised to (and received by) all the peers that should have been getting it.
This was seen on an XMR-16 that had been upgraded from 5.2.0 to 5.4.0 just a
week or so prior.

----------------------------------------------------------------------
Jon Lewis, MCP :) | I route
| therefore you are
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

Re: show ip[v6] bgp ? [ In reply to ]

sraymond at acedatacenter

Nov 11, 2013, 7:40 AM

Post #2 of 4 (1543 views)

Permalink

On Nov 11, 2013, at 7:54 AM, Jon Lewis wrote:

> "Route is advertised to X peers:" is pretty clear in what it means.
>
> "Route is to be sent to Y peers:" I can't find any documentation saying what this means...and I'm curious, what does this mean?

My observation with this output seems to mean that it will send the route to the peer, after the peer is brought up. Basically a configured but currently-down peer will get sent this route.

> The above mentioned bug was that an ipv6 route for which there was a network statement and static route to null0, the show ipv6 bgp route detail output said that the route was being advertised to a subset of the peers the config should have caused it to be sent to. Those peers were not actually receiving the route. Removing and re-adding the static route to null0 caused the route to be advertised to (and received by) all the peers that should have been getting it. This was seen on an XMR-16 that had been upgraded from 5.2.0 to 5.4.0 just a week or so prior.

Am pretty sure I have seen limited similar flakiness with new/readded statics getting advertised properly with 5.4.0c, but believe it was IPv4. Are they members of a peer-group? I found that you have to go both "peer-group blah activate" and "neighbor <ipv6> activate" or it wouldn't send any routes, period (the neighbor was establishing just fine). Same wasn't true for IPv4 peer-group members.

Re: show ip[v6] bgp ? [ In reply to ]

esk-puck.nether.net at esk

Nov 11, 2013, 7:51 AM

Post #3 of 4 (1551 views)

Permalink

I believe the 'to be sent' indicates that the route has not yet been
shared with the BGP peer, possibly because the session is down. 'sh ip
bgp summary' should give you an indication of which neighbor(s) have
routes left to send.

As for your null route, we had an issue recently where the router would
not replace a route with admin distance of 255 (even with a static route
with admin distance of 1), so be sure to use 254 or less on all of your
static routes. We only have ipv4 routes, but I'm assuming ipv6 would be
affected as well.

--
Eldon Koyle
Information Technology
Utah State University
--
BOFH excuse #380:
Operators killed when huge stack of backup tapes fell over.

On Nov 11 9:54-0500, Jon Lewis wrote:
> In troubleshooting what I believe is a bug in 5.4.0d in which an
> IPv6 route claims to be advertised to peers but is not received by
> those peers, I think I've found another issue that's more of a
> feature change than bug. But I'm curious, in the output of show ip
> bgp route detail, there are two sections near the end of the
> output...
>
> "Route is advertised to X peers:" is pretty clear in what it means.
>
> "Route is to be sent to Y peers:" I can't find any documentation
> saying what this means...and I'm curious, what does this mean?
>
> The above mentioned bug was that an ipv6 route for which there was a
> network statement and static route to null0, the show ipv6 bgp route
> detail output said that the route was being advertised to a subset
> of the peers the config should have caused it to be sent to. Those
> peers were not actually receiving the route. Removing and re-adding
> the static route to null0 caused the route to be advertised to (and
> received by) all the peers that should have been getting it. This
> was seen on an XMR-16 that had been upgraded from 5.2.0 to 5.4.0
> just a week or so prior.
>
> ----------------------------------------------------------------------
> Jon Lewis, MCP :) | I route
> | therefore you are
> _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
>
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

Re: show ip[v6] bgp ? [ In reply to ]

jlewis at lewis

Nov 11, 2013, 9:37 AM

Post #4 of 4 (1542 views)

Permalink

On Mon, 11 Nov 2013, Steven Raymond wrote:

>
> On Nov 11, 2013, at 7:54 AM, Jon Lewis wrote:
>
>> "Route is advertised to X peers:" is pretty clear in what it means.
>>
>> "Route is to be sent to Y peers:" I can't find any documentation saying what this means...and I'm curious, what does this mean?
>
> My observation with this output seems to mean that it will send the
> route to the peer, after the peer is brought up. Basically a configured
> but currently-down peer will get sent this route.

That was my first guess...but it doesn't seem to be the case. These peers
are established and have been for some time.

>> The above mentioned bug was that an ipv6 route for which there was a
>> network statement and static route to null0, the show ipv6 bgp route
>> detail output said that the route was being advertised to a subset of
>> the peers the config should have caused it to be sent to. Those peers
>> were not actually receiving the route. Removing and re-adding the
>> static route to null0 caused the route to be advertised to (and
>> received by) all the peers that should have been getting it. This was
>> seen on an XMR-16 that had been upgraded from 5.2.0 to 5.4.0 just a
>> week or so prior.
>
> Am pretty sure I have seen limited similar flakiness with new/readded
> statics getting advertised properly with 5.4.0c, but believe it was
> IPv4. Are they members of a peer-group? I found that you have to go
> both "peer-group blah activate" and "neighbor <ipv6> activate" or it
> wouldn't send any routes, period (the neighbor was establishing just
> fine). Same wasn't true for IPv4 peer-group members.

The peers in question are members of peer-groups. My experience with peer
groups has been that if you configure route-maps (or probably any other
options) for the peer-group, then it's necessary to activate the
peer-group in the address-family, or your route-maps (or other options)
are not applied to members of the peer group in that address-family.
Individual members can be activated, but activating the peer-group appears
to be sufficient. i.e. a member of an activated peer-group does not have
to be individually activated in order for routes to be exchanged.

----------------------------------------------------------------------
Jon Lewis, MCP :) | I route
| therefore you are
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp