Полиномиальный анализ по irc
its not 10k processes, just 3600
not the c10k problem, the p10k problem :-)
only 2647 of them hung ....
the rest managed to exit OK
i suspect there is a O(n^2) bug
it does seem to be very very slowly recovering
its down to 2643 hung now
2640 ....
thats a lot of cpu cycles per exit() call
for a 4 cpu xeon box :)
Sounds like the old solaris fcntl bug :-)
this is truly amazing syscall latency :)
2610
its going faster now
2467 ....
1876 ...
1009 ...
hmm, stuck at 1009 :(
You should be able to measure the speed and do a polynomial analysis over it...
ahh, its moving again
999 ...
Is it really o(n^2), or even o(e^n) :-)
dunno!
While waiting you have plenty of time :-)
we can do it from the irc log ...
LOL :-)
701
290
1
yay!