Hi,
I've more-or-less completed the analysis of the problems I was having
with the protocol startup and shutdown.
The most notable discovery is that apparently sometimes my home wireless
network appears to experience extreme packet latency - several seconds.
Given that the shutdown timer is only 10 seconds, and the retransmit
interval for initial startup is 5 seconds - it's not too hard to see how
that would stress things. And this is a good thing!
I've identified a few things I need to fix...
Server-side graceful protocol shutdown needs to wait until the
current output packets have been ACKed (with a timeout).
It looks like I have a hole where a packet is sent after the
protocol is shut down
If I receive a restart packet while the protocol is restarting, it
is _probably_ the right thing to ignore it (timeouts need to be taken
into account).
Although I can make the changes for these things during the next week, I
won't be able to test them thoroughly until I return home - since two of
my test machines came with me.
One of the reasons why these didn't show up before was because I was
only testing on my wired network.
_______________________________________________
Assimilation mailing list - Discovery-Driven Monitoring
[email protected]
http://lists.community.tummy.com/cgi-bin/mailman/listinfo/assimilation
http://assimmon.org/
