🚀 go-pugleaf

Remy Lebeau (TeamB) wrote:
>"Bob Gonder"  wrote in message
>
>> If this takes an hour to manifest, (one assumes the test program is
>> somewhat consistant in tests per second), does it mean the server
>> is being a bit slow to release "disconnected" sessions?
>
>The error has to do with the numer of pending queued connections, not the
>number of established connections.  When a client connects to a server, the
>connection is placed into a queue.  When the server calls accept(), a
>connection is removed from the queue.  When the server begins listening, it
>specifies the maximum number of connections can are allowed to be queued at
>a time.  If the maximum number of clients are already queued, any further
>clients will fail to connect until the queue opens up available slots.

So then the OP's server is just a tad slower than his testing client,
such that over the course of say, a second, the server falls behind by
one connection/accept. So after an hour of the que getting deeper and
deeper, it finaly fills up. 
IOW if the server can accept/process 100 connections per second, and
the tester sends out 101....
Or if the server pauses for garbage collection, or Windows disk flush,
and falls behind....

If this is data gathering, maybe the client could increase the packet
size on the next connection after a failure.