System error: [10061] Connection refused

[Michael Steele_2]

Your network config has several issues beginning with "...1223887 connect requests dropped due to no listener...". Which signifies unnecessary applications running and listening on ports with not incoming traffic. None at all. 1223886 is a huge number that only zeros at reboot. So how long has it been since reboot?

who -b
uptime

I'm not so concerned about "...145 connect requests dropped due to full queue...". 145 is not a bad number but it should be zero. This queue size is determined by the application and the HP-UX global parameter from /dev/tcp and ndd : "max_conn_request_max". Since your current value is 4096 I'd look to the applicaiton.

Based upon the number of retransmits I'd say you've also got router issues, or cable issues, maybe a broken cable or unterminated cable, because "...1611103417 packets sent
(1722654504 bytes) retransmitted..." is not proportional.

###########################################

Use SAM to reduce ninode and nfile:

nfile 202028 - (15*NPROC+2048)
ninode 108704 - (8*NPROC+2048)

############################################

Have your rebooted yet?

[Ali Bajunaid]

Hi,

What we did until now:

Only we change tcp_conn_request_max to 6000

who -b :

system boot May 26 09:13

uptime :

8:40am up 71 days, 23:28, 19 users, load average: 0.35, 0.41, 0.30


Regards

[Ali Bajunaid]

Hi,

Is there is away to reset the Network statistic without reboot the server.


Regards

[U.SivaKumar_2]

Hi,

lanadmin will clear it

#lanadmin


LOCAL AREA NETWORK ONLINE ADMINISTRATION, Version 1.0
Tue , Aug 14,2001 11:29:22

Copyright 1994 Hewlett Packard Company.
All rights are reserved.

Test Selection mode.

lan = LAN Interface Administration
menu = Display this menu
quit = Terminate the Administration
terse = Do not display command menu
verbose = Display command menu

Enter command: lan

LAN Interface test mode. LAN Interface PPA Number = 0

clear = Clear statistics registers
display = Display LAN Interface status and statistics registers
end = End LAN Interface Administration, return to Test Selection
menu = Display this menu
ppa = PPA Number of the LAN Interface
quit = Terminate the Administration, return to shell
reset = Reset LAN Interface to execute its selftest
specific = Go to Driver specific menu

LAN Interface test mode. LAN Interface PPA Number = 0

clear = Clear statistics registers
display = Display LAN Interface status and statistics registers
end = End LAN Interface Administration, return to Test Selection
menu = Display this menu
ppa = PPA Number of the LAN Interface
quit = Terminate the Administration, return to shell
reset = Reset LAN Interface to execute its selftest
specific = Go to Driver specific menu

Enter command: clear
Clearing LAN Interface statistics registers.

regards,

U.SivaKumar

[Ali Bajunaid]

Hi,

lanadmin -> lan -> clear

Not clearing the statistic for netstat -p tcp

Regards

[U.SivaKumar_2]

Hi,

Some patching is required for lanadmin to clear netstat statistics also.

So current option is to reboot. ;-(

After your increased tcp_conn_request_max to 6000 , did the problem repeated then ?

regards,

U.SivaKumar

[Ali Bajunaid]

Hi,

Up to now the problem not occured !!!

Regards

[Zigor Buruaga]

Hi Ali,

I have also tried the suggestion from U.Sivakumar with same results as you, netstat still shows the old statistics. So I also need some patching :-)
By the way, have you checked if the socket connections are correctly closed?
Could you post the output of "netstat -a"?

Kind regards,
Zigor

[U.SivaKumar_2]

So what is the conclusion ? ;-)

regards,

U.SivaKumar

[Ali Bajunaid]

Hi,

Conclusion we will be monitoring the system and thanks for your suppourt.


Regards

[Michael Steele_2]

How should I say this... tcp_conn_request_max to 6000 is real high.

You've got applications running and dropping packets as well as far too many retransmissions which indicate physical problems within the network.

Here's are two links for additional network utilities that will help: 'lsof' and 'tcpdump'.

http://hpux.cict.fr/hppd/hpux/Sysadmin/lsof-4.64/

http://hpux.cict.fr/hppd/hpux/Networking/Admin/tcpdump-3.6.2/

[Michael Steele_2]

PS - Those lanadmin stat's aren't the same as the information in netstat. Octets vs. packets, NIC vs. server.

The only way to clear netstat is to reboot or rebuild the lan with ifconfig, i.e., up/down, plumb/unplumb.

Here's a link on lanadmin and NICs.

http://docs.hp.com/hpux/pdf/J3622-90011.pdf

[Ron Kinner]

netstat -a appears to have been cleaned up so we can't see who is talking to whom but there appear to be an awful lot of connections to the btm_3 process. Is this perhaps a homebrew program? Perhaps the writer forgot to close his connections when they were no longer needed or opens a new one each time instead of reusing the existing one.

Ron

[Abdul Majeed Lardhi]

Hi,
I am working with Mr. Ali ,

I notified from the netstat of HP-UX that the drop connection since MAY 2003 is just only 145 !!! this means we still far from the problem !!

2nd; the HP-UX host has only few TIME_WAIT status in netstat -n while the clint side NT server has thousands of TIME_WAIT !!! and few of ESTABLIESH status !!??

Do you think its DNS Configeation Problem !!??

Regards