[CUWiN-Dev] Boatload of WRAP node problems

[David Young]

On Wed, Aug 31, 2005 at 11:25:08AM -0500, Seth Price wrote:
> As I have mentioned before, I just moved into the Race and Washington  
> intersection and I am having issues with my connection. I'm not sure  
> how many of these are isolated to my node (10.0.150.205), and how  
> many are related, so I'll lump them into one email.
> 
> - wdogctl is using ~100% of my cpu time, all the time. top  

I have patched wdogctl to fix that.  This may be behind some of the
other problems.

> - node continually sends out packets. Judging by the lights on the  
> switch, some of them are broadcast, most are not. And yes, I see  
> these continually when both desktops are in Sleep mode.

Please find out what it is sending.  It is normal for it to send broadcast
UDP packets on port 9191 all of the time.  That is the routing protocol
running.

> - Node "freezes" randomly. I've seen it go for about an hour and  
> freeze, and I've seen it go overnight without freezing (only once  
> though). I say "freezes" because although it won't respond or route,  
> but I can see that it's ethernet interface is very active by the  
> blinkenlights on the switch. SSH sessions stop responding and the  
> ping to the node goes from < 1 ms to nothing. I can reset it simply  
> by pulling the plug for a second. More top and ping info is avl if  
> requested.

The best way to figure this out is to attach to the serial console.

> - My connection is extremely unreliable. It goes out for very short  
> stretches, and it goes out for very long stretches. Yesterday I left  
> "ping google.com" going for a while on my desktop to see if I could  
> come up with any interesting stats. Here is a sampling of the kinds  
> of errors that I get (watch the ICMP seq # to see where I cut and  
> pasted, generally I only get one or two dropped packets, 10% of the  
> time):

I can tell from the "destination unreachable" and "ttl exceeded" messages
that something goes wrong with the routing.  Your packets are hopping
north on Race Street before they go south again.  They are probably
coming south through your router and looping back north.

> 36 bytes from 10.0.216.146: Destination Host Unreachable

This is a node about a block north of you.

> 36 bytes from 10.22.228.254: Destination Net Unreachable

I guess the default route went away.  That is the IP on your node's
ethernet.

> - Both web connections and ssh connections to my node take about 30  
> seconds to get an initial response, then they generally work fine.  
> With intermittent pauses.

Due to wdogctl chewing CPU, I bet.

> - I occasionally get DNS entries that direct me to the chambana.net  
> server homepage. CNN.com became chambana.net yesterday, for example.

This is a misconfiguration of the chambana.net DNS server.

> - Before I got the node's wireless interface online, the RouteViz web  
> GUI showed only my node and a single horizontal green line to the  
> right. It was like my node was connected to something off screen (no  
> matter how far I zoomed out). It was rather confusing.

Probably a stub route.  I can filter those so that routeviz never
sees them.

> - Now that my node is connected to the mesh, the routeviz images are  
> 404 errors.

No idea.

Dave


-- 
David Young             OJC Technologies
dyoung at ojctech.com      Urbana, IL * (217) 278-3933