[Dxspider-support] Ping problem

[Dirk Koopman]

On Mon, 2002-07-29 at 04:26, charlie carroll wrote:
> Dirk:
> Here's an item for a rainy day when you are bored and want something to
> do!
> 
> For some time I had noticed that the ping command did not aways work. 
> Successive pings to a node in a short period of time, seconds between
> each ping, would not get an answer.  After some longer period of time, I
> would get a ping response.  Tonight I was able to log into the next node
> in a chain and use watchdbg to see what was happening.  Here's the
> scenario...

Um....

Well, there is an intrinsic problem here which puts us between the devil
and the deep blue sea. There is no unique ID for each ping. This means
that the only way of preventing misrouting cascading out of control is
to dupe check PC51s.

In a perfect network, I would not have to dupe check PC51. Unfortunately
we don't live in such a network. There are loops and it seems I *have*
to dupe check them - even though they are 'directed' (or routed) toward
a particular station through one interface only (and not broadcast a la
PC11 [dx spot]). 

There was a spate of PC51 loops before I started to dupe check them. I
can alter the timeout on the residence time of each PC51 but I can't,
safely, take the checking away. The result of this is that if you ping
again *before* you get a reply, that extra ping may get lost and the
internal ping tracking system itself may get confused, thus giving
strange "out of order" ping times.   

If you wait for the reply, it should all work reasonably well, as the
dupe check is removed when a reply is received. 

Having said that, I wonder whether there is a hole in the dupe removal
in the intervening "routing" nodes (ie the nodes between the two end
points). 

I will look at this.

Dirk G1TLH