So, I've been troubleshooting a packet loss problem on my Cox cable internet connection. For a period of about a week, my "Quality" RRD graphs showed RTT to the gateway dropping from a 10ms average to a 2-3ms average. It jumped back up after a modem reboot.
Anyway, in the course of troubleshooting, I showed that graph to a DOCSIS engineer who informed me that a RTT of 2-3ms to the gateway was practically physically impossible, you know, due to the laws of physics.. which I tend to believe.
Fast forward a bit, and that started happening again – the reported low RTT times. I double checked by logging into pfsense via ssh and pinging the gateway. Sure enough, the RTT was more like 8ms than 2ms. I looked at the apinger config and it showed the correct gateway. I then restarted apinger from the command line, and the reported RTT fell back in line with reality.
The DOCSIS engineer suggested that my firewall may have been pinging the cable modem itself, which makes sense because when the modem is first starting up, it will assign an IP in the 192.168.100.0/24 range until it finishes connecting to the CMTS. What bothers me about this is that my ping times to the modem (at 192.168.100.1) average less than 1ms (800 us to be precise) as measured by smokeping running on a linux VM behind my firewall.
I neglected to do a pcap or tcpdump to see what apinger was actually doing during the time I witnessed the aberrant behavior. If it happens again, I'll do that, but until then, any ideas? Curious more than anything.
BS RTT values in apinger reported many times. No fix, move on.
Do a query on the forum for apinger - there are multiple, multiple threads about this.. Seems that if you for whatever reason trigger a high alarm, that the numbers then go to shit.. Easiest fix is if numbers are low restart apinger.
I have not seen any details of when, if it might be fixed? Nor have I seen anything pointing out the actual cause to why high alarm then throws the numbers off. But if you look at your RRD you will prob notice it goes high before it drops low.
Be nice if they fixed it, then again not thinking its high on anyone's priority list ;) If you want realistic graphing of your icmp response to something then yeah smokeping works great..
Thanks – my bad for not doing a search. Glad to know it's a known issue.