hi, there:

first, i'd like to thank diesel developer to share this cute lib.

currently i am testing and evaluating it. i found a strange problem
when running the http.py in examples directory.

i started a simple http server by running:
python2.6 http.py

then i started httperf to do some stress test like this:
httperf --server localhost --port 8088 --num-conn 1000 --num-call 100

the test went ok. however, when i checked with netstat: netstat -np|
grep CLOSE_WAIT
i saw about 15 dead sockets like:
tcp        0      0 127.0.0.1:8088
127.0.0.1:43312             CLOSE_WAIT  21936/python2.6
tcp        0      0 127.0.0.1:8088
127.0.0.1:43376             CLOSE_WAIT  21936/python2.6
tcp        0      0 127.0.0.1:8088
127.0.0.1:43449             CLOSE_WAIT  21936/python2.6
tcp        0      0 127.0.0.1:8088
127.0.0.1:43448             CLOSE_WAIT  21936/python2.6
tcp        0      0 127.0.0.1:8088
127.0.0.1:43447             CLOSE_WAIT  21936/python2.6

if i run more tests, there will be more dead sockets.  they accumulate
and won't go away unless i kill the server.
is it a bug?

my system is CentOS 5.2. i already tweaked tcp parameters like:
net.ipv4.tcp_fin_timeout = 30
net.ipv4.tcp_keepalive_intvl = 2
net.ipv4.tcp_keepalive_probes = 2
net.ipv4.tcp_keepalive_time = 1800

can u please take a  look at this issue? thanks a lot!

roy

Reply via email to