[Xorp-users] BGP module dies and problems with import policies

[Christian Svensson]

Hello.

Sorry for an email filled with formatting, I didn't find any good solutions
to make things clear.

Kernel: Ubuntu 8.04 server (2.6.24-23-server)
Xorp: We have tested 1.4, 1.5 and 1.6 for both issues described in this
email

Scenario: http://www.mxd.nu/info/lan-man-route-net-0901.jpg
i.e 2 BGP peers and OSPF to internal network.

Problem 1:
Around 2009-01-15 18:50 the communication out from our BGP router suddenly
stopped.
Attaching logs.

Logs:
http://www.mxd.nu/router.log
http://www.mxd.nu/router.err.log

Mirror:
https://denzel.cmd.nu/~bluecommand/xorp/2009-01-15<https://denzel.cmd.nu/%7Ebluecommand/xorp/2009-01-15>

Recently this apparently happened again:

[ 2009/01/25 06:15:26 INFO xorp_bgp BGP ] Sending: Notification Packet: Hold
Timer Expired(4)
[ 2009/01/25 06:18:19  ERROR xorp_bgp:4924 XRL +635 xrl_pf_stcp.cc die ]
XrlPFSTCPSender died: Keepalive timeout
[ 2009/01/25 06:18:19  ERROR xorp_fea:4920 XRL +635 xrl_pf_stcp.cc die ]
XrlPFSTCPSender died: Keepalive timeout
[ 2009/01/25 06:18:22  ERROR xorp_rib:4921 XRL +635 xrl_pf_stcp.cc die ]
XrlPFSTCPSender died: Keepalive timeout
[ 2009/01/25 06:18:23  ERROR xorp_policy:4922 XRL +635 xrl_pf_stcp.cc die ]
XrlPFSTCPSender died: Keepalive timeout
[ 2009/01/25 06:18:26  ERROR xorp_bgp:4924 XRL +635 xrl_pf_stcp.cc die ]
XrlPFSTCPSender died: Keepalive timeout
[ 2009/01/25 06:18:26  WARNING xorp_bgp:4924 LIBXORP +468 timer.cc
expire_one ] Timer Expiry *much* later than scheduled: behind by 17.104458
seconds
[ 2009/01/25 06:18:26  WARNING xorp_bgp:4924 LIBXORP +468 timer.cc
expire_one ] Timer Expiry *much* later than scheduled: behind by 17.104535
seconds
[ 2009/01/25 06:18:26  WARNING xorp_bgp:4924 LIBXORP +468 timer.cc
expire_one ] Timer Expiry *much* later than scheduled: behind by 17.104547
seconds
[ 2009/01/25 06:18:26 INFO xorp_bgp XRL ] Sender died (protocol = "stcp",
address = "127.0.0.1:40354")
[ 2009/01/25 06:21:26  ERROR xorp_bgp:4924 XRL +338 xrl_pf_stcp.cc die ]
STCPRequestHandler died: life timer expired

Problem 2:
We are also trying to apply a special localpref value on some routes to
balance our peering:

*set protocols bgp peer "212.112.175.81" import "PREFERED_BGP_STHLM"
*We are given the response: *"210 Transport failed*" and xorp_policy dies.

Log and configuration:
http://www.mxd.nu/policy-bgp-xorp.txt

ip ro shows all routes after XORP has died (no processes left). It will not
route any traffic on restart even though ip ro | wc -l return 270 000 (i.e.
all the routes are in there).
Stopping xorp, flushing addresses and routes on the router interfaces
followed by a start of xorp "solves" this.

-- 
Christian Svensson
Command Systems
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ICSI.Berkeley.EDU/pipermail/xorp-users/attachments/20090126/f366b4e1/attachment.html