topic Re: connect requests dropped due to full queue in Operating System - HP-UX

connect requests dropped due to full queue

Eduardo_63 — Mon, 25 Jul 2005 14:44:34 GMT

Hi Friends'

I've Oracle Applications on HP-UX.
I see much TIME_WAIT stat with the port 8000 on the server in where reside Apache, also see(netstat -p tcp):

49 connect requests dropped due to full queue

This behavior is normal ?

Thanks,

Re: connect requests dropped due to full queue

RAC_1 — Mon, 25 Jul 2005 14:58:22 GMT

I think, you need to look at three different ndd settings.

tcp_con_request_max
tcp_time_wait_interval - How long stream persists in TIME_WAIT tcp_tw_cleanup_interval - TIME_WAIT timeout expiration checking interval

You can check your current settings as follows.

ndd -h tcp_con_request_max

Anil

Re: connect requests dropped due to full queue

Eduardo_63 — Mon, 25 Jul 2005 15:43:02 GMT

Hi Anil,

The values:

tcp_con_request_max =4096
tcp_time_wait_interval=60000
tcp_tw_cleanup_interval=60000

Thanks,

Re: connect requests dropped due to full queue

Eduardo_63 — Mon, 25 Jul 2005 17:37:20 GMT

Hi friend,

the netstat show :

$ netstat -an | grep tcp| awk '{print $6}'| sort | uniq -c
421 CLOSE_WAIT
130 ESTABLISHED
629 FIN_WAIT_2
79 LISTEN
1 SYN_SENT
460 TIME_WAIT

any suggestion?

Re: connect requests dropped due to full queue

rick jones — Tue, 26 Jul 2005 15:47:06 GMT

The actual length of a listen queue will be the _minimum_ of tcp_conn_request_max and whatever the application provides in its call to listen(). So, if tcp_conn_request_max is 1024 and the application calls listen(fd,128) 128 will be the depth of the listen queue.

HP-UX, and just about any other TCP stack out there, can deal with lots of TIME_WAIT without any trouble, so there should be no need to tweak anything there.