GW checking causes dropped connections



  • First of all, amazing job on the beta, this is an amazing piece of software.

    I did a test with 2.3.b.20160302.0541… I left GW checking/dpinger on for most of the day while I had a constant RDP connection open from work and experienced it frequently dropping the connection and freezing. Then I turned it off in the middle of a really busy time and immediately stopped getting any issues. I actually left it connected overnight and came in the next morning to it still connected!

    Is there something I need to tweak or change with GW monitoring or is this a known issue?

    Thanks!



  • You may need to tweak the latency and/or loss thresholds… it may be resetting the interface thinking there's loss (the interface is down) when it's still up, just going through a momentary issue.

    Also, I found with my ISP that pinging the second hop (a real router) provided much more stable results with dpinger than pinging the first hop (my ISP's cable modem termination system (CMTS)). The CMTS has more duties than simply pushing packets around... and it's set up to give those duties more priority than ICMP echo replies. If you have DOCSIS cable internet service, you might want to try the same.



  • Or ping something out on the real internet that should always be up and responding reliably - Google 8.8.8.8 and/or 8.8.4.4 for example.



  • @phil.davis:

    Or ping something out on the real internet that should always be up and responding reliably - Google 8.8.8.8 and/or 8.8.4.4 for example.

    That's my guess as to what'll keep that from being an issue, don't ping the ISP's router IP. And/or increasing the probe interval. Probably triggering some rate limiting on the ISP's side, which we've seen from at least one other here.

    dpinger doesn't do anything to the interface even if it has loss or goes down (that only updates gateway groups with multi-WAN configs).