When I do an mtr from my home to Wikipedia, the host ae20.3518.ear1.Chicago3.level3.net shows a significant packet loss:
My traceroute [v0.92]
What I suspect is happening is only ICMP packets are being lost, not every packet going through that host, since Wikipedia doesn't seem to have degraded performance. Anyone care to offer other explanations or speculation?
Brian
This is normal for many busy routers. It’s a self defence mechanism where ICMP is always low priority (and UDP echo even more so) when compared to normal TCP/UDP transit traffic. ICMP must be handled directly by the router rather than “transit” traffic which these days is handled directly by the hardware. Routing takes priority and for busy routers it’s not uncommon for ICMP to show longer response time or packet loss.
On Jul 6, 2023, at 23:57, Brian Lowe brian2@groupbcl.ca wrote:
When I do an mtr from my home to Wikipedia, the host ae20.3518.ear1.Chicago3.level3.net http://ear1.chicago3.level3.net/ shows a significant packet loss:
My traceroute [v0.92]laptop (192.168.1.1) 2023-07-06T23:29:23-0500 Keys: Help Display mode Restart statistics Order of fields quit Packets Pings Host Loss% Snt Last Avg Best Wrst
- gateway 0.0% 21 0.4 0.4 0.3 0.4
- (ip address redacted) 0.0% 21 8.8 9.0 8.0 11.6
- rc2nr-be103-1.wp.shawcable.net http://rc2nr-be103-1.wp.shawcable.net/ 0.0% 20 8.9 9.4 8.5 10.1
- 24.244.61.17 0.0% 20 9.9 9.9 8.4 12.1
- 24.244.60.173 0.0% 20 9.4 9.7 8.9 11.3
- rc2nr-be1.wp.shawcable.net http://rc2nr-be1.wp.shawcable.net/ 0.0% 20 9.9 9.4 8.4 10.2
- rc3sc-be26.wp.shawcable.net http://rc3sc-be26.wp.shawcable.net/ 0.0% 20 9.8 10.0 9.0 12.3
- rc4ec-be25-1.il.shawcable.net http://rc4ec-be25-1.il.shawcable.net/ 0.0% 20 26.2 25.4 24.1 28.2
- lag-104.ear5.Chicago2.Level3.net http://lag-104.ear5.chicago2.level3.net/ 0.0% 20 25.3 25.3 23.9 28.8
- ae20.3518.ear1.Chicago3.level3.net http://ear1.chicago3.level3.net/ 78.9% 20 26.9 25.6 24.2 26.9
- WIKIMEDIA-F.ear1.Chicago3.Level3.net http://wikimedia-f.ear1.chicago3.level3.net/ 0.0% 20 24.9 25.5 24.2 27.5
- xe-5-2-1.cr2-codfw.wikimedia.org http://xe-5-2-1.cr2-codfw.wikimedia.org/ 0.0% 20 74.3 75.2 73.5 77.8
- text-lb.codfw.wikimedia.org http://text-lb.codfw.wikimedia.org/ 0.0% 20 74.5 73.7 69.4 76.7
What I suspect is happening is only ICMP packets are being lost, not every packet going through that host, since Wikipedia doesn't seem to have degraded performance. Anyone care to offer other explanations or speculation?
Brian _______________________________________________ Roundtable mailing list Roundtable@muug.ca mailto:Roundtable@muug.ca https://muug.ca/mailman/listinfo/roundtable
And to give you an idea of the level… The routers I administer automatically start “randomly” dropping ICMP and UDP echo (destined for them) after 10k packets per second. That said, we also see on the order of 300-500k packets per second (and more) of traffic transiting the router. I would expect that Wikipedia is busier than that.
On Jul 7, 2023, at 09:34, David Milton david@dmilton.ca wrote:
This is normal for many busy routers. It’s a self defence mechanism where ICMP is always low priority (and UDP echo even more so) when compared to normal TCP/UDP transit traffic. ICMP must be handled directly by the router rather than “transit” traffic which these days is handled directly by the hardware. Routing takes priority and for busy routers it’s not uncommon for ICMP to show longer response time or packet loss.
On Jul 6, 2023, at 23:57, Brian Lowe brian2@groupbcl.ca wrote:
When I do an mtr from my home to Wikipedia, the host ae20.3518.ear1.Chicago3.level3.net http://ear1.chicago3.level3.net/ shows a significant packet loss:
My traceroute [v0.92]laptop (192.168.1.1) 2023-07-06T23:29:23-0500 Keys: Help Display mode Restart statistics Order of fields quit Packets Pings Host Loss% Snt Last Avg Best Wrst
- gateway 0.0% 21 0.4 0.4 0.3 0.4
- (ip address redacted) 0.0% 21 8.8 9.0 8.0 11.6
- rc2nr-be103-1.wp.shawcable.net http://rc2nr-be103-1.wp.shawcable.net/ 0.0% 20 8.9 9.4 8.5 10.1
- 24.244.61.17 0.0% 20 9.9 9.9 8.4 12.1
- 24.244.60.173 0.0% 20 9.4 9.7 8.9 11.3
- rc2nr-be1.wp.shawcable.net http://rc2nr-be1.wp.shawcable.net/ 0.0% 20 9.9 9.4 8.4 10.2
- rc3sc-be26.wp.shawcable.net http://rc3sc-be26.wp.shawcable.net/ 0.0% 20 9.8 10.0 9.0 12.3
- rc4ec-be25-1.il.shawcable.net http://rc4ec-be25-1.il.shawcable.net/ 0.0% 20 26.2 25.4 24.1 28.2
- lag-104.ear5.Chicago2.Level3.net http://lag-104.ear5.chicago2.level3.net/ 0.0% 20 25.3 25.3 23.9 28.8
- ae20.3518.ear1.Chicago3.level3.net http://ear1.chicago3.level3.net/ 78.9% 20 26.9 25.6 24.2 26.9
- WIKIMEDIA-F.ear1.Chicago3.Level3.net http://wikimedia-f.ear1.chicago3.level3.net/ 0.0% 20 24.9 25.5 24.2 27.5
- xe-5-2-1.cr2-codfw.wikimedia.org http://xe-5-2-1.cr2-codfw.wikimedia.org/ 0.0% 20 74.3 75.2 73.5 77.8
- text-lb.codfw.wikimedia.org http://text-lb.codfw.wikimedia.org/ 0.0% 20 74.5 73.7 69.4 76.7
What I suspect is happening is only ICMP packets are being lost, not every packet going through that host, since Wikipedia doesn't seem to have degraded performance. Anyone care to offer other explanations or speculation?
Brian _______________________________________________ Roundtable mailing list Roundtable@muug.ca mailto:Roundtable@muug.ca https://muug.ca/mailman/listinfo/roundtable
On 2023-07-06 Brian Lowe wrote:
When I do an mtr from my home to Wikipedia, the host ae20.3518.ear1.Chicago3.level3.net shows a significant packet loss:
You both nailed it. If a hop is dropping but the hops after are lossless, then it's just that hop doing weird stuff with ICMP. Some hops might just blanket drop ICMP, depending on their config. No law says you have to respond to it.
-
Brian Lowe -
David Milton -
Trevor Cordes