BGP Session drops when loading routes into a VRF

Question

I've configured a small MPLS network with 4 P routers (ASR 1002s) 4 PE routers (6509s) and two Cisco 7206s as route reflectors.  I'm using OSPF as the routing protocol on the PE-CE interfaces and have sham links configured between PE routers.  I currently have two VRFs configured on this network, one is working fine, and there are approximately 150 routes in this VRF. 
The second vrf is configured, it also has ospf sham links configured, neighbors up, etc.
The problem I have is that when I bring up the first PE-CE link and load about 7000 routes into the second vrf my BGP sessions between the other PEs and route-reflectors start timing out.  I do show the vpnv4 routes show up in the route-reflectors "sh ip bgp vpnv4 rd 14017:2" I believe the debug outputs below indicate the issue is in the route-reflectors, but was curious if anyone else had seen this issue.  The BGP peering IP address for one of the route-reflectors is 10.2.0.7 and one of the session dropping PEs is 10.2.0.13.
Normal output from "unloaded" debug ip bgp vpnv4 unicast keepalives:
Route-reflector output:
May 16 09:35:31   2329: May 16 09:35:30.718 CDT: BGP: ses global 10.2.0.13 (0xA0519C0:1) Keep alive timer fired.
May 16 09:35:31   2330: May 16 09:35:30.718 CDT: BGP: 10.2.0.13 KEEPALIVE requested (bgp_keepalive_timer_expired)
May 16 09:35:31   2331: May 16 09:35:30.718 CDT: BGP: ses global 10.2.0.13 (0xA0519C0:1) service keepalive IO request.
May 16 09:35:31   2332: May 16 09:35:30.718 CDT: BGP: 10.2.0.13 KEEPALIVE write request serviced in BGP_IO
PE output:
May 16 09:35:38.421 CDT: BGP: 10.2.0.7 KEEPALIVE requested (bgp_keepalive_timer_expired)
May 16 09:35:38.421 CDT: BGP: 10.2.0.7 KEEPALIVE sent
Route-reflector output:
May 16 09:35:39   2341: May 16 09:35:38.427 CDT: BGP: 10.2.0.13 received KEEPALIVE, length (excl. header) 0
Same output but during a "loaded" condition:
Route-reflector output:
May 15 20:41:31   774: May 15 20:41:31.015 CDT: BGP: ses global 10.2.0.13 (0xA091324:1) Keep alive timer fired.
May 15 20:41:31   775: May 15 20:41:31.015 CDT: BGP: 10.2.0.13 KEEPALIVE requested (bgp_keepalive_timer_expired)
May 15 20:41:31   778: May 15 20:41:31.015 CDT: BGP: ses global 10.2.0.13 (0xA091324:1) service keepalive IO request.
May 15 20:42:29   793: May 15 20:42:28.363 CDT: BGP: ses global 10.2.0.13 (0xA091324:1) Keep alive timer fired.
May 15 20:42:29   794: May 15 20:42:28.363 CDT: BGP: 10.2.0.13 KEEPALIVE requested (bgp_keepalive_timer_expired)
May 15 20:43:23   805: May 15 20:43:22.638 CDT: BGP: ses global 10.2.0.13 (0xA091324:1) Keep alive timer fired.
May 15 20:43:23   806: May 15 20:43:22.638 CDT: BGP: 10.2.0.13 KEEPALIVE requested (bgp_keepalive_timer_expired)
May 15 20:43:33   813: May 15 20:43:33.934 CDT: %BGP-3-NOTIFICATION: received from neighbor 10.2.0.13 4/0 (hold time expired) 0 bytes
PE output:
May 15 20:43:33.927 CDT: %BGP-3-NOTIFICATION: sent to neighbor 10.2.0.7 4/0 (hold time expired) 0 bytes
May 15 20:43:33.927 CDT: %BGP-5-ADJCHANGE: neighbor 10.2.0.7 Down BGP Notification sent
Route-reflector output:
May 15 20:43:34   814: May 15 20:43:33.934 CDT: %BGP-5-ADJCHANGE: neighbor 10.2.0.13 Down BGP Notification received
It appears to me that we are missing "BGP: 10.2.0.13 KEEPALIVE write request serviced in BGP_IO"
For full disclosure there are other BGP keep-alive events going on to the other PEs at this time, some passing, some failing.
route-reflector info:
route-reflector1#sh inv
NAME: "Chassis", DESCR: "Cisco 7206VXR, 6-slot chassis"
PID: CISCO7206VXR      , VID:    , SN: 37050753
NAME: "NPE-G2 0", DESCR: "Cisco 7200 Series Network Processing Engine NPE-G2"
PID: NPE-G2            , VID: V03 , SN: JAF1410AADM
NAME: "disk2", DESCR: "256MB Compact Flash Disk for NPE-G2"
PID: MEM-NPE-G2-FLD256 , VID:    , SN:
NAME: "Power Supply 1", DESCR: "Cisco 7200 AC Power Supply"
PID: PWR-7200-AC       , VID:    , SN:
NAME: "Power Supply 2", DESCR: "Cisco 7200 AC Power Supply"
PID: PWR-7200-AC       , VID:    , SN:
route-reflector1#sh ver
Cisco IOS Software, 7200 Software (C7200P-ADVIPSERVICESK9-M), Version 15.1(4)M5, RELEASE SOFTWARE (fc1)
Technical Support: http://www.cisco.com/techsupport
Copyright (c) 1986-2012 by Cisco Systems, Inc.
Compiled Tue 04-Sep-12 19:41 by prod_rel_team
ROM: System Bootstrap, Version 12.4(12.2r)T, RELEASE SOFTWARE (fc1)
BOOTLDR: Cisco IOS Software, 7200 Software (C7200P-BOOT-M), Version 15.1(4)M5, RELEASE SOFTWARE (fc1)
route-reflector1 uptime is 15 weeks, 20 hours, 43 minutes
System returned to ROM by reload at 11:03:48 UTC Thu Jan 31 2013
System restarted at 12:37:52 CST Thu Jan 31 2013
System image file is "disk2:c7200p-advipservicesk9-mz.151-4.M5.bin"
Any thoughts on this problem would be greatly appreciated.

Answer 1

Hi again,
did you check this?
http://www.cisco.com/en/US/tech/tk365/technologies_configuration_example09186a008010a28a.shtml
I honestly do not remember the defaults in Cisco but I had a very similar issue working with Juniper because of this value was set @ about 4500 prefixes at once.
Take care
Alessio
Sent from Cisco Technical Support iPad App