On 2023-04-19 13:46, Per Jessen wrote:
Carlos E. R. wrote:
Isengard:~ # while sleep 1 ; do DATE=`date --iso=s` ; echo -n "$DATE " ; fping -c 20 --period=100 --quiet 2a02:...80d4 ; done
2023-04-19T10:11:07+02:00 2a02:...:80d4 : xmt/rcv/%loss = 20/16/20%, min/avg/max = 0.45/0.53/0.69 2023-04-19T10:11:10+02:00 2a02:...:80d4 : xmt/rcv/%loss = 20/0/100% 2023-04-19T10:11:13+02:00 2a02:...:80d4 : xmt/rcv/%loss = 20/11/45%, min/avg/max = 0.47/1.09/5.45
What do you make of 100%& loss?
Even at 10 pings/second, that is surely not right. Some questions are -
a) do the echo requests reach your router?
AFAIK, yes. Once I connected the laptop to the switch_1, with everything mirrored to that port, running ethereal. AFAIK there were pings coming from upstairs that were not answered.
b) does your router respond with echo replies?
Not all.
c) does isengard receive the echo reply?
Those that are sent, yes, AFAIK.
telcontar, same period
2023-04-19T10:10:01+02:00 router : xmt/rcv/%loss = 100/86/14%, min/avg/max = 0.27/0.39/4.83 2023-04-19T10:10:12+02:00 router : xmt/rcv/%loss = 100/92/8%, min/avg/max = 0.22/0.38/4.82 2023-04-19T10:10:23+02:00 router : xmt/rcv/%loss = 100/88/12%, min/avg/max = 0.27/0.39/4.13
10% unanswered pings still seems pretty high, even at 10/sec, but way better. What is the difference between isengard and telcontar?
They were not running the exact same command at time. For that, I have to leave the commands running while, telcontar is not running a constant ping.
Isengard:~ # while sleep 1 ; do DATE=`date --iso=s` ; echo -n "$DATE " ; fping -c 100 --quiet router ; done 2023-04-19T14:03:57+02:00 router : xmt/rcv/%loss = 100/79/21%, min/avg/max = 0.26/0.39/1.04 2023-04-19T14:05:38+02:00 router : xmt/rcv/%loss = 100/85/15%, min/avg/max = 0.27/0.39/1.09 2023-04-19T14:07:19+02:00 router : xmt/rcv/%loss = 100/84/16%, min/avg/max = 0.26/0.40/1.53 2023-04-19T14:09:00+02:00 router : xmt/rcv/%loss = 100/85/15%, min/avg/max = 0.24/0.42/0.84 2023-04-19T14:10:41+02:00 router : xmt/rcv/%loss = 100/83/17%, min/avg/max = 0.24/0.70/25.1 2023-04-19T14:12:22+02:00 router : xmt/rcv/%loss = 100/84/16%, min/avg/max = 0.26/0.35/0.66 2023-04-19T14:14:03+02:00 router : xmt/rcv/%loss = 100/80/20%, min/avg/max = 0.28/0.43/2.15 2023-04-19T14:15:44+02:00
Telcontar:~ # while sleep 1 ; do DATE=`date --iso=s` ; echo -n "$DATE " ; fping -c 100 --quiet router ; done 2023-04-19T14:03:54+02:00 router : xmt/rcv/%loss = 100/84/16%, min/avg/max = 0.25/0.73/26.9 2023-04-19T14:05:35+02:00 router : xmt/rcv/%loss = 100/89/11%, min/avg/max = 0.27/0.33/1.30 2023-04-19T14:07:16+02:00 router : xmt/rcv/%loss = 100/90/10%, min/avg/max = 0.25/0.51/7.74 2023-04-19T14:08:58+02:00 router : xmt/rcv/%loss = 100/90/10%, min/avg/max = 0.24/0.82/22.9 2023-04-19T14:10:39+02:00 router : xmt/rcv/%loss = 100/87/13%, min/avg/max = 0.25/1.00/24.9 2023-04-19T14:12:20+02:00 router : xmt/rcv/%loss = 100/89/11%, min/avg/max = 0.26/1.51/47.3 2023-04-19T14:14:01+02:00 router : xmt/rcv/%loss = 100/83/17%, min/avg/max = 0.20/0.85/45.5 2023-04-19T14:15:42+02:00
They look similar to me. A bit higher in Isengard, perhaps.z Both are machine upstairs, same switch. Telcontar is the desktop machine, AMD, powerful. Isengard is the mini server, Intel, MSI Cubi N MiniPC. After lunch I'll have more data.
I think I would try some large download, maybe an ISO file and then keep an eye on the interface stats. Run wireshark and look for tcp errors. Something like that.
I'll buy that. But I don't know if this will show the errors between router and switch:
router--1--SW1--2--SW2--3---telcontar
The problem is in "1" when going to SW2.
I assume 1, 2 and 3 are cables/connections? physical or wireless?
Ethernet cable.
Your pings from telcontar (above) traverse both switches, yet "performance" is better. I presume isengard is connected to sw1 ?
No, both to sw2, upstairs. This moment I have both machines running the same fping command; we have to wait some time to collect data and compare.
in black:
I don't do html :-)
The mail was sent plain. I refer to "black" inside ethereal. Packets are coloured.
The capture file is 600 MB, if you want it.
Even if I did, I would want the raw file.
-rw------- 1 root root 708638880 Apr 19 10:25 capture_iso_download.pcapng I have still not closed ethereal, so I can still save the capture in any other format ;-)
A lot of "TCP Out-of-Order" would make me curious.
In the very beginning, I also see a couple of interesting entries:
16 10:24:53.327894931 192.168.1.14 192.168.1.16 ICMP 125 Destination unreachable (Port unreachable)
18 10:24:53.327937181 192.168.1.14 1.1.1.1 ICMP 125 Destination unreachable (Port unreachable)
20 10:24:53.328056641 192.168.1.14 1.0.0.1 ICMP 125 Destination unreachable (Port unreachable)
Yes, I pasted some that made me curious. I could also download same file from Isengard down to a laptop on sw1 and run a capture. I say Isengard because it has apache, the laptop doesn't. To do reverse direction, has to be different protocol, like sftp. Also connect laptop directly on router, and compare both captures. Those captures should not have private information if run on the new laptop, so I could share easily. -- Cheers / Saludos, Carlos E. R. (from 15.4 x86_64 at Telcontar)