Forum - Extreme packet loss in sixxs graphs :: SixXS

Extreme packet loss in sixxs graphs

Shadow Hawkins on Friday, 02 July 2004 15:32:56

I checked around a little by modifying the URL of the loss graph and the graphs I found seemed to have just a few % loss now and then. My loss graph is showing an AVERAGE of 70%! I have no trouble using IPv6 though, and pings to noc.sixxs.net and my gateway (2001:960:2:29d::1, Scarlet POP) go through with 0% packet loss. Anybody have any idea what's going on? I'm a bit afraid I'm gonna lose my account because of that graph. How is that data collected btw, so I can try a bit for myself? I use Linux 2.6.7, using heartbeat. Commands to set the tunnel up are the same as in the guide so nothing wrong there. The graph pic: http://www.ipv4.sixxs.net/home/tunnelinfo/latency/?type=loss&graphtime=1w&pop=nlams04&ip=2001:960:2:29d::2 Edit: --- 2001:960:2:29d::1 ping statistics --- 100 packets transmitted, 100 received, 0% packet loss, time 99086ms --- noc.sixxs.net ping statistics --- 100 packets transmitted, 100 received, 0% packet loss, time 103223ms

Extreme packet loss in sixxs graphs

Carmen Sandiego on Saturday, 03 July 2004 00:13:05

Well probably your have to many downloads so your connection is filled up. So that's the reason probably

Extreme packet loss in sixxs graphs

Shadow Hawkins on Saturday, 03 July 2004 10:59:54

Nope unfortunately not, I'm barely using my connection now and I definately didn't use it at all 4-5 hours ago, yet I had 70% loss all friday. :{

Extreme packet loss in sixxs graphs

Carmen Sandiego on Saturday, 03 July 2004 14:17:42

Hmm... What connection you have? maybe you schould choose tunnel closer to you... which latency would be smaller

Extreme packet loss in sixxs graphs

Shadow Hawkins on Saturday, 03 July 2004 20:48:19

Latency isn't a problem either, I got 40 average, topped at 60 for a few minutes yesterday but nothing else. I'm using the Scarlet POP which I had the best latency to of all the POPs when I registered. My connection is 8Mbps downstream and 800kbps upstream (in theory ofc, I get about 6750/640), ADSL. Edit: More stats. An answer to how the data for the graphs is collected would be really nice, I'm clueless. jarjar@exscape.org jarjar $ ping -c 100 nlams04.sixxs.net | grep -A 3 'statistics' &&

ping6 -c 100 nlams04.sixxs.net | grep -A 3 'statistics'

--- nlams04.sixxs.net ping statistics --- (IPv4) 100 packets transmitted, 100 received, 0% packet loss, time 99087ms rtt min/avg/max/mdev = 36.110/36.713/37.434/0.307 ms --- nlams04.sixxs.net ping statistics --- (IPv6) 100 packets transmitted, 100 received, 0% packet loss, time 99085ms rtt min/avg/max/mdev = 36.445/37.139/37.901/0.374 ms jarjar@exscape.org jarjar $ ping -c 100 noc.sixxs.net | grep -A 3 'statistics' &&

ping6 -c 100 noc.sixxs.net | grep -A 3 'statistics'

--- noc.sixxs.net ping statistics --- (IPv4) 100 packets transmitted, 100 received, 0% packet loss, time 99087ms rtt min/avg/max/mdev = 64.641/65.347/67.647/0.531 ms --- noc.sixxs.net ping statistics --- (IPv6) 100 packets transmitted, 100 received, 0% packet loss, time 98987ms rtt min/avg/max/mdev = 41.252/41.769/42.633/0.346 ms 0% loss, at the same time my graph shows about 90%(!).

Extreme packet loss in sixxs graphs

Carmen Sandiego on Sunday, 04 July 2004 00:03:08

If you do a tcpdump on your tunnel, do you see a icmp-reply for every incoming icmp-echo request? Are you sure you are not mentioned in the FAQ ([url=https://noc.sixxs.net/faq/connectivity/?faq=conntracking]My tunnel goes down after some idletime. My tunnelendpoint also is a NAT[/url])?

Extreme packet loss in sixxs graphs

Shadow Hawkins on Sunday, 04 July 2004 23:09:03

Yes, I have read the FAQ and made appropriate changes (2 days ago). Besides, it doesn't seem like the tunnel is down at any time. I just did a tcpdump on my sixxs interface and eth0. All pings that I sent get a reply and same is true about the pings from broker04.ams.nl.sixxs.net. I also tried to ping that host myself (regular ICMPv4), 0% packet loss. This is weird. Anyway, I upped my ICMPv4 limit from 1 ping per second (should be enough, since I haven't gotten any other pings at this time), to 5 per second. Edit: In case you ask, the limit is 300 per minute, not literally 5 per second. So I *can* receive 100 ICMPs the very same second if necessary. I started a tcpdump on both eth0 (ipv4) and sixxs (ipv6), I have got less than 20 ICMP packets in total (+ reply packets) in 5 minutes so the limits can't be the problem either I guess. UPDATE: It seems the problem was my firewall script after all - DUH! Even after setting it to allowing ALL icmp unlimited, it still forced a damn limit somewhere. I'll start writing my own iptables script tomorrow I guess.

Extreme packet loss in sixxs graphs

Jeroen Massar SixXS Staff

on Tuesday, 06 July 2004 10:28:54

Guess what, SixXS pings using ICMPv6 as mentioned in the FAQ (Endpoint doesn't ping).

Please note Posting is only allowed when you are logged in.