On 3 Jan 2007 09:26:27 -0800,
wrote:
>When I do the repair I have discovered what is fixing the problem, it
>is the clearing of the arp cache.
>
>I know this because I did a manual clear of the arp cache with "arp -d
>*" and that fixed it.
>
>I dont know much about arp but hopefully someone here can make some
>conclusions about this?
Weird. You didn't answer my question as to whether your wireless ISP
is part of a mesh network. If so, it would make sense that the
destination gateway might change depending on the topology of the
moment. I'll keep things simple and avoid this possibility.
The obvious question is whether the MAC address of the gateway
(192.168.252.254) changes before it disconnects, and after it
recovers. Try it before and after and see if there's a change. If it
does change, well then you'll need to do something to forcibly expire
the arp cache and ping the gateway, which should renew the entry.
However, if it doesn't change, then try this experiment. Run:
arp -s 192.168.252.254 xx-xx-xx-xx-xx-xx
to permanently set the MAC address of the gateway. If this fixes it,
my guess(tm) is that either your ethernet driver or IP stack on your
computah is having a bad day. It's suppose to send an ARP request to
the gateway immediately after it detects a connection. It's not.
I just tried to simulate your problem. I have an ancient DWL-900AP+
setup in client mode connected to the neighbors WRT54G. Encryption is
off. When I disconnect the antenna to simulate a connection loss, it
takes about 2 minutes for XP to recognize that the connection is gone.
Various services (AIM, Skype, PPTP VPN) fail prior to XP announcing a
lost connection.
When I put the antenna back and try to ping the neighbors router, it
takes about 20 seconds to re-establish the connection. Yours
apparently takes either much longer or never succeeds. I just did it
again, but this time, I had a continuously running FPING session
running. The reconnection was about 5 seconds. This is the way it
should work.
>> Explain what equipment. Make, model and exact ways it's all wired together.
>
>Its a Trango Fox 5310 Subscriber Unit. The ethernet cable goes from the
>back of the PC into a Power over Ethernet box, out of there and up to
>the roof where the Trango is. No routers at my end.
Nice:
<http://www.trangobroadband.com/products/fox5310_international.shtml>
5.3GHz. No mesh network. No microwave oven interference. Do you
have line of sight? How far away is the central access point.
>Yep - take a look at a screenshot here:
>http://img54.imageshack.us/my.php?image=ss1nr5.jpg
Perfect. No problems with the IP setup. I can't seem to get the
secondary DNS server to respond to my DNS queries, but it might be
firewalled to accept queries only from the WISP's network.
nslookup
Default Server: dns1.snfcca.sbcglobal.net
Address: 206.13.28.12
> server 201.55.12.1
Default Server: [201.55.12.1]
Address: 201.55.12.1
> www.cruzio.com
Server: [201.55.12.1]
Address: 201.55.12.1
DNS request timed out.
timeout was 2 seconds.
DNS request timed out.
timeout was 2 seconds.
*** Request to [201.55.12.1] timed-out
This is probably unrelated to the current problem, but you should
check if your secondary DNS server is functional from your end.
>It always sends a successful ping to the gateway when working
>(occasionaly 1% loss)
Are the ping times (latency in msec) constant? In other words, do
they always show the same number of msec, or do they vary all over the
place? If they vary, it's a sign of interference or possibly wireless
congestion. The extra delays are signs of packet retransmissions.
Unfortunately, I can't tell where the gateway IP is located in your
WISP's network, so it's difficult to isolate just your traffic
results.
>I did try to ping the gateway when the connection when down, and on my
>second attempt I got a return, initally I thought it was just the
>connection between me the gateway being fixed, but now I think the
>whole connection was repaired. Sometimes it fixes itself after a few
>minutes - sometimes it takes longer. Have not quite worked this out
>exactely yet.
It requires traffic to fix itself. If the arp cache is being flushed,
it won't repopulate the ARP cache untill it sends something to the
gateway. What I find interesting is that the gateway appears to be
functional, but nothing beyond it. Try using traceroute (Windoze
tracert) to something else is going down along the path. It might be
the backhaul between the central access point and where it hits a
wired connection.
--
# Jeff Liebermann 150 Felker St #D Santa Cruz CA 95060
# 831-336-2558
#
http://802.11junk.com
#
http://www.LearnByDestroying.com AE6KS