I've been racking my brain for two days now, searching Cisco documentation, my CCNP study materials and performing various web searches and I've yet to find a cause for this odd issue.
I work for a Internet Service Provider and for testing purposes while we wait for native IPv6 backbone connectivity I've been setting up IPv6 via OSPF on our core network with a /32 we've been assigned by ARIN and testing external connectivity through a ipv6ip tunnel with Sprint set up on our Sprint backbone router. We have a BGP peering session running over the tunnel with Sprint.
We are experiencing uniform packet loss (80 percent) when pinging from routers routing through the backbone router over the tunnel to the IPv6 internet. However, when we ping from the backbone router through the tunnel to the IPv6 internet we experience no packet loss. Furthermore we can ping from our routers to the backbone router tunnel address just fine.
The backbone router is a Cisco 12K running GRP-B management and the adjacent routers/switches are Cisco 6509s running Sup720 supervisor cards.
Here is some ping examples:
The backbone router can ping ipv6.google.com just fine through the tunnel, which means the tunnel and BGP is running fine.
router.backbone.metroix#ping ipv6 ipv6.google.com
Translating "ipv6.google.com"...domain server (184.108.40.206) [OK]
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 2001:4860:B004::68, timeout is 2 seconds:
Success rate is 100 percent (5/5), round-trip min/avg/max = 220/221/224 ms
We can ping from a adjacent router to the backbone router's local tunnel address fine:
switch.core.metroix#ping ipv6 2001:440:EEEE:FF34::2
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 2001:440:EEEE:FF34::2, timeout is 2 seconds:
Success rate is 100 percent (5/5), round-trip min/avg/max = 0/0/0 ms
But when we ping from the adjacent router across the tunnel to the address on the other side or any ipv6 internet address we get the following:
switch.core.metroix#ping 2001:440:EEEE:FF34::1 repeat 20
Sending 10, 100-byte ICMP Echos to 2001:4860:B004::68, timeout is 2 seconds:
Packet sent with a source address of 2607:FD80:FFFF::1
Reply to request 0 (184 ms)
Request 1 received unknown echo response type U
Request 2 timed out
Request 3 received unknown echo response type U
Request 4 timed out
Reply to request 5 (184 ms)
Request 6 received unknown echo response type U
Request 7 timed out
Request 8 received unknown echo response type U
Request 9 timed out
Success rate is 20 percent (2/10), round-trip min/avg/max = 184/184/184 ms
The pattern of response, U, drop, U, drop, repeat does not change and packet loss to any IPv6 site is 80 percent.
We've statically assigned the adjacent routers with default routes pointing to the backbone router and the backbone router itself has a statically assigned default route pointing to the tunnel.
Here is the current configuration and default route information:
switch.core.metroix#sho ipv6 route ::/0
Routing entry for ::/0
Known via "static", distance 1, metric 0
Redistributing via ospf 10
Route count is 1/1, share count 0
2607:FD80:FFFF::3 <--- Our backbone router
Last updated 15:19:06 ago
(From our backbone router)
router.backbone.metroix#sho ipv6 route ::/0
S ::/0 [1/0]
via ::, Tunnel1
This is our tunnel configuration:
description IPv6 TO SPRINT
no ip address
no ip directed-broadcast
ipv6 address 2001:440:EEEE:FF34::2/64
tunnel source Loopback1
tunnel destination (censored)
tunnel mode ipv6ip
On the Cisco 12K backbone router, The OC-3 Uplink to Sprint is an engine 0 Card and the connection to the adjacent 6509s is on an engine 1 gig Ethernet card.