Apache 1.3.18 on HP-UX 9000

[serguey kavafian]

We are getting "[Mon Feb 24 14:56:13 2003] [error] (233)No buffer space available: accept: (client socket)" in Apache error_log.
Where could be we look for solution?

[Francois Bariselle_3]

Hi,

Does it exist a message in /var/adm/syslog.

Frank.

[Steven E. Protter]

Okay you are a little behind on version number, and maybe you didn't use hp depots to install.

Try this maybe:
https://payment.ecommerce.hp.com/cgi-bin/swdepot_parser.cgi/cgi/try.pl?productNumber=B9415AAPA1327&date=

Next, or as an alternative run the command as root ipcs.

Check out the output. A properly running apache is going to use a lot of shared memory resources.

Might want to redo the kernel making the message queues larger, shmmax, parameters like that.

This could be a resource issue.

Has apache ever worked properly on the box?

If not, look into the December 2002 patch, then relink apache if you self compiled it.

Among these steps, one of them might hook you up. I'd look at shared memory. That's what has me thinking right now.

SEP

[serguey kavafian]

Thank you all for the answers!

This version of the Apache HTTP server was as a result of Oracle 8i database server installation. It is working fine but there are some of these errors in the error_log that I wanted to know the reason/solution.

We installed a new Apache version 1.3.26 from the soft depot for the static content (html, docs and etc).

[W.C. Epperson · Accepted solution]

This is system error 233. errno.h tells us this is ENOBUFS condition. When a web server is squawking this error, the TCP listen queue usually needs tuning. One page with some guidance is at:
http://www.hp.com/products1/unix/webservers/apache/techtips/hpapache_20_perf.htm

[serguey kavafian]

Thank you W.C. Epperson for the replay.
I also thought that this should be some dynamic setting parameter, because it happened during pick hours.
Will monitor the result.