Page MenuHomeVyOS Platform
Create Task
Maniphest T984

BGP received routes not installed to FIB
Closed, ResolvedPublicBUG
Actions

Description

There has been a regression somewhere between 1.1.8 and 1.2.0-rc6 where learned routes from an eBGP peer are not installed to the FIB.

With the following configuration:

bgp 397143 {
    neighbor 169.254.169.254 {
        address-family {
            ipv4-unicast {
                nexthop-self
                soft-reconfiguration {
                    inbound
                }
            }
        }
        ebgp-multihop 2
        remote-as 64515
    }
    parameters {
        router-id 23.157.160.2
    }
}

No learned routes are displayed in ip route show.

However in 1.1.8, with the following configuration:

bgp 397143 {
    neighbor 169.254.169.254 {
        ebgp-multihop 2
        nexthop-self
        remote-as 64515
        soft-reconfiguration {
            inbound
        }
    }
    parameters {
        router-id 23.157.160.2
    }
}

All learned routes appear in ip route show.

This ticket may be related to T982.

Details

Difficulty level
Unknown (require assessment)
Version
1.2.0-rc6
Why the issue appeared?
Will be filled on close

Related Objects

Event Timeline

bswinnerton created this task.Nov 8 2018, 10:45 PM
bswinnerton updated the task description. (Show Details)
fromport added a subscriber: fromport.Nov 8 2018, 10:56 PM
syncer triaged this task as Normal priority.Nov 9 2018, 12:49 PM
syncer added a project: VyOS 1.2 Crux (VyOS 1.2.0-rc7).
syncer added subscribers: dmbaturin, syncer.Nov 13 2018, 3:47 AM

@dmbaturin

pstaszewski [3:13 AM]
i just found some prolem with 6.1-dev
when sequence start is
bgpd -> zebra
fib is not filled with all learned routes from bgpd (edited)
sequence start neeed to be
zebra sleep 1sec bgpd
here is how it looks like when i start first bgpd then zebra (local routes interfaces is 349)
ip r | wc -l
349
after 5 mins
ip r | wc -l
349
and here how it looks like when i start zebra - sleep 1 sec and after bgpd
ip r | wc -l
349
after few swcs
ip r | wc -l
4002
after 10 secs
ip r | wc -l
11401

after 20 secs
ip r | wc -l
25035
after 40 secs
ip r | wc -l
142473

syncer edited projects, added VyOS 1.2 Crux (VyOS 1.2.0-rc8); removed VyOS 1.2 Crux (VyOS 1.2.0-rc7).Nov 13 2018, 3:48 AM
bswinnerton added a comment.EditedNov 13 2018, 3:59 AM

A user in Slack today pointed out that this only happens with V4 prefixes, V6 prefixes are properly installed into the FIB.

bswinnerton added a comment.Nov 18 2018, 4:37 AM

I'm not positive, but it seems that this only occurs when ebgp-multihop is enabled on the peer.

syncer edited projects, added VyOS 1.2 Crux (VyOS 1.2.0-rc9); removed VyOS 1.2 Crux (VyOS 1.2.0-rc8).Nov 20 2018, 12:44 PM
syncer edited projects, added VyOS 1.2 Crux (VyOS 1.2.0-rc10); removed VyOS 1.2 Crux (VyOS 1.2.0-rc9).Nov 27 2018, 12:03 AM
bswinnerton added a comment.EditedDec 5 2018, 1:58 PM

I ended up moving away from VyOS but the more that I think this problem, I wonder if it's due to mutihop being on and not having a route to the next hop.

In my example above, it may be worth trying something like set protocols static route 169.254.169.254/32 next-hop 10.0.0.1.

sebastianm added a subscriber: sebastianm.Dec 5 2018, 6:29 PM

adding a static route towards the vultr gateway fixes this, as @bswinnerton pointed out.

bswinnerton added a comment.Dec 5 2018, 6:35 PM

👍 cool. Since we've confirmed that as a solution, I think it's safe to close.

@sebastianm, if you believe that disable-connected-check is not working properly, it may be worth opening another ticket.

bswinnerton closed this task as Resolved.Dec 5 2018, 6:35 PM
syncer added a project: VyOS-1.2.0-GA.Dec 5 2018, 9:32 PM
syncer moved this task from Needs Triage to Finished on the VyOS 1.2 Crux (VyOS 1.2.0-rc10) board.