<p dir="ltr"><br>
On Feb 10, 2015 10:58 AM, "R. G. Newbury" <<a href="mailto:newbury@mandamus.org">newbury@mandamus.org</a>> wrote:</p>
<p dir="ltr">> I don't think that the problem really relates to the DNS server. I think it has something to do with how the cableco's gateway initializes the new link from your router, to the rest of the network (a port has been closed, and is not re-opened???)<br>
><br>
> Some thing seems to be lost such that your cableco/ISP's setup does not immediately recognize the need for a DNS lookup of the SD IP address or does not know where to request the IP address.<br>
><br>
> But things wake up when ping comes calling!<br>
> But also note that ping uses ICMP, NOT tcp or udp.<br>
><br></p>
<p dir="ltr">What this sounds like is a TCP connection being left open without any data sent, for long enough that a NAT 'forgets' the connection is still alive.</p>
<p dir="ltr">Perhaps Rogers is starting to use some form of carrier-grade-nat implementation primed for when they run out of ipv4 allotment, and ARIN says 'no'. Although, this is unlikely, as the time out is on a wholly new connection and not one I would expect to be kept open and idle like ssh.</p>
<p dir="ltr">Another potential is that perhaps DNS from their server actually is overloaded and times out for a short period during which the script may also timeout? I imagine most people configure their servers to use NTP... The synchronization of numerous mythtv boxes all connecting probably looks like a mini-ddos on the DNS and web services both...</p>
<p dir="ltr">If you're able to get a packet capture with tcpdump when this happens, it may be able to [dis]prove some theories.</p>
<p dir="ltr">Mike</p>