Summary: TCP/IP settings

From: <>
Date: Mon Jun 03 2002 - 08:07:29 EDT
Thanks everyone.



Try a netstat -i to see if you have any interface errors. Check to see if
there are errors or a duplex speed mismatch by checking these settings on
the switch that you are connected to. The error your getting is a TCP error
but if the lower levels of networking aren't configured proeprly it could
cause timeots and a hit on performance. By correcting speed/duplex
mismatches on my switch i got our sites down from 5 second pageloads to 1.25
and collectively they get more than 1 million hits a month.

There are some tuning hints for websphere on Solaris at:


I would definitely consider exploring setting your device to hard-coded 100 
full-duplex if possible.  Talk with your network administrator to make sure 
you are plugged into a switch which is capable of full-duplex mode, then run 
through the following:

# ndd -set /dev/hme adv_10fdx_cap 0                <-- don't use 10 full 
# ndd -set /dev/hme adv_10hdx_cap 0                <-- don't use 10 half 
# ndd -set /dev/hme adv_100hdx_cap 0          <-- don't use 100 half duplex
# ndd -set /dev/hme adv_100fdx_cap 1      <-- DO use 100 full duplex
# ndd -set /dev/hme adv_autoneg_cap 0    <-- don't use auto-negotiation

Note that the last ndd above will cause the hme interface to reset, causing a 
temporary network outage.  You'll also want to make sure that the interface 
on the switch is hard-set to 100/full as well, auto-negotiation is bad on 
either end.

Also, did you make sure to set the tcp_conn_req_max_q on system startup?  
Parameters set with ndd will not persist across reboot, they must be set on 
startup every time (typically via an initialization script under /etc/rc*.d).
sunmanagers mailing list
Received on Mon Jun 3 08:12:51 2002

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:42:45 EST