On 2023-04-18 21:43, Per Jessen wrote:
Carlos E. R. wrote:
To avoid having packets dropped. ICMPs may be dropped, they have the lowest priority. If you are flood pinging your network - which I suggest 100 pings per second is - some will be dropped.
The test takes a minute to run. I tried 25 mS, no difference. Any command that does ping tests fails the test.
The '-i' argument seems to set the time between pings to different machines, try using '-p' instead.
Runs faster.
The same test, if run directly, without a switch, works perfectly.
On a ptp link, I'm sure it does.
It proves that the router has no problem responding a ping flood.
The old router worked fine. They replaced the router, and I lost ALL connectivity switch-router. They replaced the router, and I have this, 30% loss. When I try any web site, there is a delay, compared to the old router. Many packages fail, but being TCP it tries again.
FWIW, between two office machines on the same Netgear Gbit switch, I think it is only shared with a printer - I ran
fping --quiet -6 -p1 -c1000 - 0% loss.
Okay, with your invocation, it does a ping per second, which you would think ought to be fine - still, from my TW test machine to our main firewall, at 1ping/sec, I also see 10% loss. (I think it traverses maybe three switches, I'm not sure).
ICMPs can be ignored, just like UDP.
Fine, suggest some other way to prove/test connectivity from computer to router. I think I also tested with some other protocol instead of ICMP. Again: many web pages take seconds to start responding and load. -- Cheers / Saludos, Carlos E. R. (from 15.4 x86_64 at Telcontar)