Mailing List Archive

Quagga 1.1.1 problems - config file not read correctly(?) and bgpd restarts frequently
Hello,

I am trying to upgrade from Quagga 0.99.24 -> 1.1.1 and am running into the
same two problems consistently. I'm not sure how to diagnose or fix either
problem.

1) The bgpd configuration isn't working when I first start bgpd. I'm using
the same configuration file that I use for 0.99.24. When bgpd first starts
it looks like it knows that it is using bgp and what its neighbor is, but
the routes are not being is advertised.

If I start vtysh, remove the bgp configuration with "no router bgp <asn>",
then cut and paste the commands from bgpd.conf into the terminal then
advertising will kick up immediately.

Here is my configuration: (NOTE: I've anonymized it by changing the first
octet of ipv4 to 9 or 10, and the fist three numbers of the ASNs to 100)

---
log syslog debugging
log record-priority
log timestamp precision 3

debug bgp events
debug bgp filters
debug bgp fsm
debug bgp keepalives
debug bgp updates

router bgp 10033
bgp router-id 9.6.10.204
network 10.6.33.159/32 route-map metric
network 10.6.33.195/32 route-map metric
network 10.6.32.243/32 route-map metric

neighbor 9.6.10.193 remote-as 10093
neighbor 9.6.10.193 update-source 9.6.10.204
neighbor 9.6.10.193 soft-reconfiguration inbound
neighbor 9.6.10.193 prefix-list policy-in in
neighbor 9.6.10.193 prefix-list policy-out out

ip prefix-list policy-in seq 5 permit 0.0.0.0/0
ip prefix-list policy-in seq 10 deny any
ip prefix-list policy-out seq 5 permit 10.6.33.159/32
ip prefix-list policy-out seq 6 permit 10.6.33.195/32
ip prefix-list policy-out seq 7 permit 10.6.32.243/32
ip prefix-list policy-out seq 10 deny any

route-map metric permit 5
set metric 400
-----

2) bgpd is restarting every four minutes or so. So once I manually bring up
advertising with vtysh it will last until bgpd shuts down and our process
monitor restarts it. Once restarted it is the same state from #1 where none
of the routes are advertised until I remove and re-add the configuration
manually with vtysh. I had enabled extra debugging and caught the shutdown
in the logs: (NOTE: I've anonymized the logs by changing the first octet of
ipv4 to 10)

2017-02-28T14:00:00.023399-05:00 my_server bgpd[45246]: 10.6.66.193
KEEPALIVE rcvd
2017-02-28T14:00:00.982301-05:00 my_server bgpd[45246]: 10.6.66.193 [FSM]
Timer (routeadv timer expire)
2017-02-28T14:00:09.274474-05:00 my_server bgpd[45246]: Terminating on
signal
2017-02-28T14:00:09.274510-05:00 my_server bgpd[45246]: %NOTIFICATION: sent
to neighbor 10.6.66.193 6/3 (Cease/
Peer Unconfigured) 0 bytes
2017-02-28T14:00:09.274518-05:00 my_server bgpd[45246]: Notification sent
to neighbor 10.6.66.193: type 6/3
2017-02-28T14:00:09.274695-05:00 my_server bgpd[45246]: %NOTIFICATION: sent
to neighbor 10.6.66.193 6/3 (Cease/
Peer Unconfigured) 0 bytes
2017-02-28T14:00:09.274704-05:00 my_server bgpd[45246]: Notification sent
to neighbor 10.6.66.193: type 6/3
2017-02-28T14:00:09.274736-05:00 my_server bgpd[45246]: %NOTIFICATION: sent
to neighbor 10.6.66.193 6/2 (Cease/
Administratively Shutdown) 0 bytes
2017-02-28T14:00:09.274743-05:00 my_server bgpd[45246]: Notification sent
to neighbor 10.6.66.193: shutdown
2017-02-28T14:00:09.274760-05:00 my_server bgpd[45246]: %ADJCHANGE:
neighbor 10.6.66.193 Down Neighbor deleted
2017-02-28T14:00:09.275391-05:00 my_server bgpd[45246]: VRF 0 is to be
deleted.
2017-02-28T14:00:09.275400-05:00 my_server bgpd[45246]: VRF 0 is to be
disabled.
Re: Quagga 1.1.1 problems - config file not read correctly(?) and bgpd restarts frequently [ In reply to ]
Someone asked me directly what OS I was using, the response was - The OS is
Scientific Linux 7.1 - though I am building Quagga 1.1.1 myself using the
latest Fedora Core RPM spec as a starting point. The Quagga build included
with SL7 is 0.99.22 which is two minor revisions less then our Quagga
0.99.24 build which we use on Scientific Linux 6.5 servers. As long as I'm
rebuilding Quagga for SL7 I'm trying to bring in the more modern version. Our
build uses the options from the Fedora Core spec except for disabling SNMP,
enabling PAM, and setting --enable-multipath=64.

Later I did a build of Quagga 0.99.24 for SL7 using the same spec file I
used for Quaga 1.1.1 (with only version changed) and it started right up
and began advertising. I think that eliminates everything except Quagga
itself.

I'll try a couple pre-1.1.1 versions to see if I can spot roughly when it
stopped working.


On Tue, Feb 28, 2017 at 2:44 PM, CJ Ess <zxcvbn4038@gmail.com> wrote:

> Hello,
>
> I am trying to upgrade from Quagga 0.99.24 -> 1.1.1 and am running into
> the same two problems consistently. I'm not sure how to diagnose or fix
> either problem.
>
> 1) The bgpd configuration isn't working when I first start bgpd. I'm using
> the same configuration file that I use for 0.99.24. When bgpd first starts
> it looks like it knows that it is using bgp and what its neighbor is, but
> the routes are not being is advertised.
>
> If I start vtysh, remove the bgp configuration with "no router bgp <asn>",
> then cut and paste the commands from bgpd.conf into the terminal then
> advertising will kick up immediately.
>
> Here is my configuration: (NOTE: I've anonymized it by changing the first
> octet of ipv4 to 9 or 10, and the fist three numbers of the ASNs to 100)
>
> ---
> log syslog debugging
> log record-priority
> log timestamp precision 3
>
> debug bgp events
> debug bgp filters
> debug bgp fsm
> debug bgp keepalives
> debug bgp updates
>
> router bgp 10033
> bgp router-id 9.6.10.204
> network 10.6.33.159/32 route-map metric
> network 10.6.33.195/32 route-map metric
> network 10.6.32.243/32 route-map metric
>
> neighbor 9.6.10.193 remote-as 10093
> neighbor 9.6.10.193 update-source 9.6.10.204
> neighbor 9.6.10.193 soft-reconfiguration inbound
> neighbor 9.6.10.193 prefix-list policy-in in
> neighbor 9.6.10.193 prefix-list policy-out out
>
> ip prefix-list policy-in seq 5 permit 0.0.0.0/0
> ip prefix-list policy-in seq 10 deny any
> ip prefix-list policy-out seq 5 permit 10.6.33.159/32
> ip prefix-list policy-out seq 6 permit 10.6.33.195/32
> ip prefix-list policy-out seq 7 permit 10.6.32.243/32
> ip prefix-list policy-out seq 10 deny any
>
> route-map metric permit 5
> set metric 400
> -----
>
> 2) bgpd is restarting every four minutes or so. So once I manually bring
> up advertising with vtysh it will last until bgpd shuts down and our
> process monitor restarts it. Once restarted it is the same state from #1
> where none of the routes are advertised until I remove and re-add the
> configuration manually with vtysh. I had enabled extra debugging and caught
> the shutdown in the logs: (NOTE: I've anonymized the logs by changing the
> first octet of ipv4 to 10)
>
> 2017-02-28T14:00:00.023399-05:00 my_server bgpd[45246]: 10.6.66.193
> KEEPALIVE rcvd
> 2017-02-28T14:00:00.982301-05:00 my_server bgpd[45246]: 10.6.66.193 [FSM]
> Timer (routeadv timer expire)
> 2017-02-28T14:00:09.274474-05:00 my_server bgpd[45246]: Terminating on
> signal
> 2017-02-28T14:00:09.274510-05:00 my_server bgpd[45246]: %NOTIFICATION:
> sent to neighbor 10.6.66.193 6/3 (Cease/
> Peer Unconfigured) 0 bytes
> 2017-02-28T14:00:09.274518-05:00 my_server bgpd[45246]: Notification sent
> to neighbor 10.6.66.193: type 6/3
> 2017-02-28T14:00:09.274695-05:00 my_server bgpd[45246]: %NOTIFICATION:
> sent to neighbor 10.6.66.193 6/3 (Cease/
> Peer Unconfigured) 0 bytes
> 2017-02-28T14:00:09.274704-05:00 my_server bgpd[45246]: Notification sent
> to neighbor 10.6.66.193: type 6/3
> 2017-02-28T14:00:09.274736-05:00 my_server bgpd[45246]: %NOTIFICATION:
> sent to neighbor 10.6.66.193 6/2 (Cease/
> Administratively Shutdown) 0 bytes
> 2017-02-28T14:00:09.274743-05:00 my_server bgpd[45246]: Notification sent
> to neighbor 10.6.66.193: shutdown
> 2017-02-28T14:00:09.274760-05:00 my_server bgpd[45246]: %ADJCHANGE:
> neighbor 10.6.66.193 Down Neighbor deleted
> 2017-02-28T14:00:09.275391-05:00 my_server bgpd[45246]: VRF 0 is to be
> deleted.
> 2017-02-28T14:00:09.275400-05:00 my_server bgpd[45246]: VRF 0 is to be
> disabled.
>
>