From: Stout, Noelette (NStout@IKON.com)
Date: Thu Feb 06 2003 - 09:04:20 EST
My fellow admins,
I have been wrestling with a problem for over a year now. I have searched
the archives, found similiar questions asked, but no solutions/summaries
have been posted. My apologies to those of you who may find this
repetitive.
My production web servers ar being load balanced by a pair of F5 Big-IP load
balancers. The F5's each poll each of the web servers every 10 seconds. Our
web servers ar E420R's running Solaris 8 last patched at the end of December
with the current cluster at that time. Periodically (about every 2 weeks
give or take a few days), the tcp connections from the F5's will hang in a
CLOSE_WAIT state. Once this happens, all of the connections from the F5's
will start to hang in CLOSE_WAIT. It is only a matter of minutes before port
80 becomes so flooded it is no longer responsive to incoming requests.
I have spoken with Sun/iPlanet support about this. I had already done
everything which they recommended to do as far as tuning Solaris and
iPlanet; they had no further advice or help to offer. Since I know some of
you will ask, the tcp_time_wait_interval (formerly tcp_close_wait_interval)
is set to 1500, which is the absolute lowest setting Sun recommends.
We have come to a complete standstill on this issue with all vendors
claiming that the problem is not on their end. I will appreciate any and
all advice and suggestions on where to go from here. It's becoming a bit
annoying to me that my Sun boxes have to be rebooted more often than the
Windows servers :-(
Thanks in advance for all your help!
Noelette Stout
Unix System Administrator
IKON Office Solutions, Inc
tel. (678)728-0611
fax (678)547-2192
email: nstout@ikon.com
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:25:46 EDT