#642 closed request (fixed)
some IPv6 not working from Dover/Tennyson
Reported by: | Owned by: | ||
---|---|---|---|
Priority: | tbd | Milestone: | ietf-088 |
Component: | incoming | Keywords: | |
Cc: | My Current Location: | ||
My MAC Address: | My OS: |
Description
my work desktop: dooku-[~] mcr 5555 %host knothole.gatineau.credil.org knothole.gatineau.credil.org has address 10.10.5.40 knothole.gatineau.credil.org has IPv6 address 2001:4830:116e:1:224:21ff:fea7:5e27 my publically visible server (same IPv6 subnet, sixxs tunnel) dooku-[~] mcr 5556 %host orlando.gatineau.credil.org orlando.gatineau.credil.org has address 132.213.238.235 orlando.gatineau.credil.org has IPv6 address 2001:4830:116e:1::212 Despite what I said in person, the problem is not wired vs wireless, but seems to have something to with the lower /64. Note that I don't get past the third hop, so it's not a problem with the LAN there. Reverse traceroute: orlando-[~] mcr 8121 %mtr -6 -r -n 2001:67c:370:128:21a:6bff:fe6a:a96a HOST: orlando.gatineau.credil.org Loss% Snt Last Avg Best Wrst StDev 1. 2001:4830:116e:1::1 0.0% 10 0.6 1.0 0.6 4.9 1.4 2. 2001:4830:1100:154::1 0.0% 10 31.0 30.8 30.7 31.0 0.1 3. 2001:4830:e1:b::2 0.0% 10 30.5 30.8 30.5 31.4 0.3 4. 2001:4830:e1:b::1 0.0% 10 31.0 30.9 30.6 31.2 0.2 5. 2001:4830:ff:b100::1 0.0% 10 30.9 31.3 30.7 32.3 0.5 6. 2001:4830:ff:b101::1 0.0% 10 36.4 36.8 36.4 37.4 0.3 7. 2001:4830:ff:f150::2 0.0% 10 44.2 44.2 43.8 44.5 0.2 8. 2001:4830:ff:1234::2 0.0% 10 102.3 102.3 101.8 103.4 0.5 9. 2001:4830:ff:1001::1 0.0% 10 112.9 113.1 112.6 114.0 0.5 10. 2001:4830:ff:f201::1 0.0% 10 144.2 144.6 144.2 145.3 0.4 11. 2001:4830:ff:f200::2 0.0% 10 145.5 144.8 144.1 145.5 0.4 12. 2001:504:16::354 0.0% 10 145.7 145.1 144.5 145.9 0.5 13. ??? 100.0 10 0.0 0.0 0.0 0.0 0.0 my laptop on ietf-1: dooku-[~] mcr 5551 %ifconfig wlan0 wlan0 Link encap:Ethernet HWaddr 00:19:d2:c0:9f:1e inet addr:31.133.176.72 Bcast:31.133.183.255 Mask:255.255.248.0 inet6 addr: 2001:67c:370:176:219:d2ff:fec0:9f1e/64 Scope:Global inet6 addr: fe80::219:d2ff:fec0:9f1e/64 Scope:Link inet6 addr: 2001:67c:370:176:a4de:d671:c818:8e5f/64 Scope:Global dooku-[~] mcr 5553 %mtr -n -r -6 knothole.gatineau.credil.org HOST: dooku Loss% Snt Last Avg Best Wrst StDev 1.|-- 2001:67c:370:176::2 0.0% 10 4.8 3.4 0.7 6.5 1.9 2.|-- 2001:569:fff::a 0.0% 10 41.4 25.5 2.5 109.3 32.0 3.|-- 2001:568:1::50d 0.0% 10 6.4 9.1 5.3 12.5 2.4 4.|-- 2001:504:16::7577 0.0% 10 6.4 10.0 6.4 17.2 3.4 5.|-- 2001:4830:ff:f200::1 0.0% 10 9.1 8.5 4.7 11.8 1.9 6.|-- 2001:4830:ff:f201::2 0.0% 10 40.5 39.9 36.4 45.0 2.6 7.|-- 2001:4830:ff:1001::2 0.0% 10 49.6 51.5 48.0 55.9 2.6 8.|-- 2001:4830:ff:1234::1 0.0% 10 106.3 109.5 106.3 115.1 2.5 9.|-- 2001:4830:ff:f150::1 0.0% 10 120.1 116.8 113.4 129.2 4.8 10.|-- 2001:4830:ff:b101::2 0.0% 10 119.5 123.0 119.5 130.5 3.5 11.|-- 2001:4830:ff:b100::2 0.0% 10 120.7 122.9 119.4 130.7 4.3 12.|-- 2001:4830:e1:b::2 0.0% 10 124.2 122.2 120.0 125.0 1.8 13.|-- 2001:4830:1100:154::1 0.0% 10 119.7 123.1 119.0 145.5 8.0 14.|-- 2001:4830:116e:1:224:21ff 10.0% 10 153.3 151.3 149.3 153.5 1.6 dooku-[~] mcr 5554 %mtr -n -r -6 orlando.gatineau.credil.org HOST: dooku Loss% Snt Last Avg Best Wrst StDev 1.|-- 2001:67c:370:176::2 0.0% 10 1.9 3.8 0.7 7.9 2.8 2.|-- 2001:569:fff::a 0.0% 10 1.3 18.7 1.3 59.6 20.2 3.|-- ??? 100.0 10 0.0 0.0 0.0 0.0 0.0 WIRED: my laptop on wired: dooku-[~] mcr 5542 %ifconfig eth0 eth0 Link encap:Ethernet HWaddr 00:1a:6b:6a:a9:6a inet addr:31.133.132.29 Bcast:31.133.135.255 Mask:255.255.248.0 inet6 addr: 2001:67c:370:128:21a:6bff:fe6a:a96a/64 Scope:Global inet6 addr: fe80::21a:6bff:fe6a:a96a/64 Scope:Link dooku-[~] mcr 5559 %mtr -n -r -6 orlando.gatineau.credil.org HOST: dooku Loss% Snt Last Avg Best Wrst StDev 1.|-- 2001:67c:370:128::3 0.0% 10 0.3 53.4 0.3 531.7 168.0 2.|-- 2001:67c:370:228::2 0.0% 10 0.3 0.3 0.3 0.4 0.0 3.|-- 2001:569:fff::a 0.0% 10 1.0 10.6 0.7 69.4 21.5 4.|-- ??? 100.0 10 0.0 0.0 0.0 0.0 0.0 dooku-[~] mcr 5560 %mtr -n -r -6 knothole.gatineau.credil.org HOST: dooku Loss% Snt Last Avg Best Wrst StDev 1.|-- 2001:67c:370:128::3 0.0% 10 0.3 0.3 0.2 0.4 0.0 2.|-- 2001:67c:370:228::2 0.0% 10 0.3 0.3 0.2 0.6 0.1 3.|-- 2001:569:fff::a 0.0% 10 1.1 15.3 0.9 61.0 20.2 4.|-- 2001:568:1::50d 0.0% 10 4.1 4.6 4.0 8.6 1.4 5.|-- 2001:504:16::7577 0.0% 10 5.1 5.0 5.0 5.1 0.0 6.|-- 2001:4830:ff:f200::1 0.0% 10 5.0 5.1 5.0 5.2 0.1 7.|-- 2001:4830:ff:f201::2 0.0% 10 36.5 36.6 36.4 37.1 0.2 8.|-- 2001:4830:ff:1001::2 0.0% 10 47.1 47.3 47.0 48.2 0.3 9.|-- 2001:4830:ff:1234::1 0.0% 10 105.2 105.7 105.1 107.3 0.6 10.|-- 2001:4830:ff:f150::1 0.0% 10 112.0 112.4 111.6 113.1 0.5 11.|-- 2001:4830:ff:b101::2 0.0% 10 117.6 118.1 117.3 119.5 0.6 12.|-- 2001:4830:ff:b100::2 0.0% 10 121.2 119.1 118.1 121.2 0.9 13.|-- 2001:4830:e1:b::2 10.0% 10 118.6 118.7 118.2 119.4 0.4 14.|-- 2001:4830:1100:154::1 0.0% 10 118.3 118.3 117.7 120.0 0.7 15.|-- 2001:4830:116e:1:224:21ff 10.0% 10 147.7 148.5 147.7 150.8 1.0
Attachments (1)
Change history (10)
Changed 7 years ago by
Attachment: | untitled-part.bin added |
---|
comment:1 Changed 7 years ago by
Owner: | changed from llynch@… to cdoyle@… |
---|---|
Status: | new → assigned |
comment:2 Changed 7 years ago by
We are currently working with our upstream provider to isolate and resolve a v6 routing through a transit AS.
comment:3 Changed 7 years ago by
I want to mention that things are more functional, or least more deterministic: I find that either all addresses work, or none of them. Hitting up-arrow return enough times makes things work.
(I was using my aiccu/sixxs tunnel for awhile to work around this)
comment:4 Changed 7 years ago by
There is clearly something bad going on; after the first hop in the Telus network we see 50% loss on some IPv6 paths.
My working theory is that there's ECMP but one of the paths is not actually forwarding IPv6 properly. This is one of the failure modes that can happen more easily when transporting IPv6 prefixes over IPv4 transport, because the BGP session doesn't share state with the IPv6. Of course, I have no idea what the IPv6 topology is inside Telus or if there's meant to be ECMP, it is just something that fits the 50% loss we see.
To Arista in Santa Clara (Level3):
traceroute6 to 2001:550:401::401 (2001:550:401::401) from 2001:67c:370:229::12, 64 hops max, 12 byte packets 1 rtra 0.239 ms 0.415 ms 0.189 ms 2 2001:569:fff::a 7.237 ms 0.830 ms 0.870 ms 3 2001:568:1::50d 4.025 ms * 4.863 ms 4 * 2001:550:3::2e1 4.277 ms * 5 2001:550::67:1113 25.087 ms 25.431 ms * 6 2001:550::25 24.914 ms * 25.078 ms 7 * 2001:550::1040 25.100 ms * 8 2001:550:2:3::8:1 25.094 ms * 25.215 ms 9 * 2001:550:2:3::8:2 24.288 ms * 10 2001:550:401::401 25.025 ms * *
When we hit this bad path, we start getting nearly exactly every other packet lost.
comment:5 Changed 7 years ago by
The pings at
https://management.meeting.ietf.org/smokeping/smokeping.cgi?target=Links.External-IPv6.arista
and
https://management.meeting.ietf.org/smokeping/smokeping.cgi?target=Links.External-IPv6.mcr
show 100% loss. If my theory holds, this is because the pings are hashing to the half of the ECMP that's not working.
comment:6 Changed 7 years ago by
Telus has fixed an issue inside their network, and all of my tests are passing now. mcr, want to re-test?
comment:7 Changed 7 years ago by
Resolution: | → fixed |
---|---|
Status: | assigned → closed |
Telus identified a router issue. This issue has been resolved and all seems well in the world of v6.
comment:8 Changed 7 years ago by
All the connections that I could not make now seem to work.
Thanks for pursuing this.
Added by email2trac