I have three Netgear Routers: WNDR3700, WNDR4000, and WNDR4500.
On the SAME wired network, the 3700 works great. But both the 4000 and 4500 exhibit the same symptom. They periodically drop WAN connections for a minute (almost exactly), then allow them for 5-20 seconds, drop them for another minute, etc. In the plots below, red in the time graphs means that traceroutes and pings are not getting returned by the affected hops along the path (top part of the plots).
Existing connections (Netflix streaming) seem to not be affected. I see this periodic dropping clearly doing traceroutes or pings. Pingplotter is especially effective at showing this.
This is what a normal Pingplotter connection (on the WNDR3700) looks like:
It is normal for the destination (the Comcast nameserver) to refuse to answer some packets. The white area on each plot is the jitter, which increases up to a loss, and then drops.
What could be causing this?
Snoopytime on the Netgear Forum suggested:
Try going to Setup, then to WAN Setup . . then UNCHECK 'IGMP Proxy Disable'.
I think I figured it out. Pingplotter sends multiple ICMP requests. The router apparently has trouble dealing with these when queued up.
I modified Pingplotter to send them serially, and the problem went away.
So what is causing it? Well, someone on the forum said that what happens is that the router loses its DNS addresses. This would be consistent with my symptoms during the periodic outages: existing connections continue to work, new ones do not. The person suggested that I put in the fixed Comcast DNS addresses (220.127.116.11 and 18.104.22.168). But I also installed the latest Netgear beta firmware which he said supposedly fixed things. And it has been working using the unmodified Pingplotter for 14 hours now.
(c) 2011 by James A. Rome. All rights reserved.