On Sun, 20 Aug 2000, James Cameron wrote:
> quozl    25701  0.1  1.9  5084 3788 ?        S    Aug15   9:08 ./netrekd
> quozl    25701  0.1  2.1  5344 4048 ?        S    Aug15  10:22 ./netrekd
> 
> Note the VSZ and RSS increase.  Note also the CPU time.
> So even a metaserver query is likely to not help the problem.

If I do the same thing, 345 connections to the list port:
USER       PID %CPU %MEM  SIZE   RSS TTY STAT START   TIME COMMAND
xyzzy     5706  0.8  1.1  1184   560  ?  S    04:32   0:01 netrekd 
xyzzy     5706  1.3  1.1  1184   560  ?  S    04:32   0:02 netrekd 


> I'm not using the default .ports file.  But I am using 2596.  Recall it
> is a fixed size array though, so the remaining zero entries are okay.
> 
> Here is the .ports file, less comments;
> 
> 2592 ./ntserv "ntserv"
> 2591 ./tools/players "players" m
> 2593 ./ntserv "ntservobs" -q 6
> 2596 special statistics
> 
> Your point about 2596 being missing may be important.

Your .ports file is broken.  You need to have quotes around the 3rd word, but
statistics has no quotes.  This causes the parser to decided that the line is
bad, and ignore it.  If you look at the variable num_progs in the scope of
main(), you'll see that it is only 3 and not 4.  This shouldn't cause
netrekd any problems, but you might want to fix your file.

Also, do you have a .nocount file?  It needs to be in LIBDIR and be readable
by the netrekd process.  Otherwise netrekd will spawn /usr/ucb/netstat on each
connection, yuck.  IMHO, a .count file to turn on this feature would make more
sense.