[Outages-discussion] [outages] Spectrum/Charter IPv6 connectivity to CloudFlare down in Texas

Jeremy Chadwick jdc at koitsu.org
Thu Apr 20 01:22:41 EDT 2023


(Moving to -discussion because there's nothing concrete in what I'm
about to say, i.e. nothing worth waking people up for.)

1. Your mtrs have lost their monospace/whitespace formatting for
whatever reason, making them difficult to read.

2. Your IPv6 mtr looks like it indicates POSSIBLE packet loss starting
at hop #4 and seems to trickle down through hops #5/6/7.  I can't be
100% certain the loss is real (it could be ICMP deprio on Charter's
routers, but only for IPv6?!) because I have not seen what a working v6
mtr from you which uses the same routers/path.  If it IS real loss, then
the problem is much more likely within Spectrum/Charter's network, but I
can't say that with absolute certainty, leading me to:

2. You have no return-path mtr for either v4 or v6, so you only are able
to see 50% of the entire picture.  Asymmetric routing is commonplace
these days, which is why you need it from both directions.

If you really and absolutely think this is a CloudFlare problem (I am
not sold on that given the above two points), then we continue with:

3. Based on 141.101.74.104 in your v4 mtr, it looks like
Spectrum/Charter has a peering agreement with CloudFlare.

4. CloudFlare does not offer a public looking glass, but they do offer
an HTTP API-based traceroute, though it is sub-par in quality and kind
of bizarre, not to mention I don't know if it supports v6:
https://blog.clucas.fr/2021/11/18/traceroute-from-cloudflare-without-looking-glass/

Given items #3/4, have you tried reaching out to CloudFlare for a
return-path mtr back to whatever your v6 source IP is?

Their contact information is public: https://www.peeringdb.com/net/4224

-- 
| Jeremy Chadwick                              jdc_at_koitsu.org |
| UNIX Systems Administrator                      PGP 0x2A389531 |
| Making life hard for others since 1977.                        |

On Wed, Apr 19, 2023 at 11:37:10PM -0400, Daniel Marks via Outages wrote:
> Have been dealing with this for 2 days, nobody at Spectrum can figure this out. 
> 
> Daniel Marks
> May Mobility, Inc
> 
> > 
> > On Apr 19, 2023, at 19:26, Carl Perry via Outages <outages at outages.org> wrote:
> > 
> > Things have been rather slow the last couple of days, just figured out why; No IPv6 connectivity to any CloudFlare services.
> > 
> > > mtr -6 --report-wide cloudflare.com
> > Start: 2023-04-19T23:20:42+0000
> > HOST: defiant Loss%   Snt   Last   Avg  Best  Wrst StDev
> >   1.|-- 2603-8081-2340-2868-0215-17ff-fe94-6457.res6.spectrum.com 0.0%    10    3.6  19.1   1.5  93.2  36.1
> >   2.|-- 2603-90c5-0003-0180-0000-0000-0000-0001.inf6.spectrum.com 0.0%    10   14.5  36.6  11.3 171.2  50.0
> >   3.|-- lag-63.hcr02rdrktxaz.netops.charter.com 0.0%    10   18.8  37.3  12.0 143.9  45.3
> >   4.|-- lag-19.ausxtxir02r.netops.charter.com 60.0%    10   65.8  30.2  11.7  65.8  24.9
> >   5.|-- lag-22.rcr01hstqtx02.netops.charter.com 80.0%    10   26.5  25.5  24.5  26.5   1.4
> >   6.|-- lag-16.hstqtx0209w-bcr00.netops.charter.com 90.0%    10   25.7  25.7  25.7  25.7   0.0
> >   7.|-- lag-12.dllstx976iw-bcr00.netops.charter.com 40.0%    10   25.4  26.3  18.2  44.4   9.5
> >   8.|-- ??? 100.0    10    0.0   0.0   0.0   0.0   0.0
> > 
> > IPv4 works however:
> > 
> > > mtr -4 --report-wide cloudflare.com
> > Start: 2023-04-19T23:20:21+0000
> > HOST: defiant                                     Loss%   Snt Last   Avg  Best  Wrst StDev
> >   1.|-- _gateway                                     0.0%    10 2.1   1.8   1.4   2.1   0.2
> >   2.|-- cpe-70-114-160-1.austin.res.rr.com           0.0%    10 51.9  34.4  14.3 149.7  42.1
> >   3.|-- lag-63.hcr01rdrktxaz.netops.charter.com      0.0%    10 13.1  38.5  13.1 159.1  44.6
> >   4.|-- lag-19.ausutxla01r.netops.charter.com        0.0%    10 18.1  21.0  14.3  34.8   6.0
> >   5.|-- lag-22.rcr01dllatx37.netops.charter.com      0.0%    10 23.4  23.3  18.3  33.3   4.3
> >   6.|-- lag-14.dllstx976iw-bcr00.netops.charter.com  0.0%    10 19.8  27.0  16.1  51.2  11.9
> >   7.|-- lag-0.pr3.dfw10.netops.charter.com           0.0%    10 39.3  28.5  14.1  62.3  16.7
> >   8.|-- 141.101.74.104                              40.0%    10 53.2  44.3  19.4 120.2  39.2
> >   9.|-- 172.71.168.2                                 0.0%    10 28.3  28.8  10.8  79.4  19.5
> >  10.|-- 104.16.133.229                               0.0%    10 122.0  42.8  16.9 122.0  34.4
> > 
> > It's kind of wild how many things don't fall back to IPv4 quickly (lots of Android TV apps) or at all (.NET). I tried reaching out to Spectrum support, but being a lowly consumer I was only told that "The modem doesn't have any filters" and that "the problem may be on the CloudFlare side", which was sadly what I predicted. So, I figured I would raise the issue here in the hopes that actual network people would see it.
> > 
> >   -Carl
> > 
> > 
> > _______________________________________________
> > Outages mailing list
> > Outages at outages.org
> > https://www.google.com/url?q=https://puck.nether.net/mailman/listinfo/outages&source=gmail-imap&ust=1682551587000000&usg=AOvVaw0omaxNXXnf3Ff4dHAQDhA8
> _______________________________________________
> Outages mailing list
> Outages at outages.org
> https://puck.nether.net/mailman/listinfo/outages


More information about the Outages-discussion mailing list