httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Siler, Vernon R." <>
Subject RE: [BUG]: "child process doesn't accept http request after proxy request" on Solaris 2.x
Date Tue, 25 Feb 1997 20:38:28 GMT
I tried removing the proxy for gopher and wais but the problem
is still there.  How would I check the mod_status?  Is that the
status code available in the log files?

I discovered after the bug report that the problem is not 
happening on Solaris 2.5.  Unfortunately our WWW server 
will be on Solaris 2.4 for the forseeable future.  

I've done some testing at the code level and managed to 
find one line that can be commented out that apparently 
fixes our problem.  My testing was mostly empirical and I'm
unclear on exactly what the code is doing.

The first thing I discovered was that our problem was fixed
if the socket used to receive the data remotely was never 
closed. (variable sock in proxy_http_handler)

In trying to figure out why that would be the case
I discoved that if line 226 of proxy_http.c in version 
1.2b7 is commented out the problem goes away.  The line is:

note_cleanups_for_fd(pool, sock);

I'm completely unclear on why this would be a problem though
I did notice that in non-error situations this command is
performed again in the bpushfd call several lines below.

Also, the actual error occurs in the saferead function in
buff.c.  The read call fails with error number 77 (EBADMSG)

Is this a bug that will be fixed or just a quirk with Solaris
2.4, our machine, our compiler, etc?  Will running the server
with this line commented out cause any problems?  

Thanks for any help you can provide. 
>From: 	Chuck Murcko[]
>Sent: 	Monday, February 24, 1997 11:48 PM
>To: 	Siler, Vernon R.
>Subject: 	Re: [BUG]: "child process doesn't accept http request after proxy
>request" on Solaris 2.x
>Hmmm. The Apache proxy doesn't support gopher or wais at present, so
>that could be causing some odd behavior.
>ProxyRemote currently needs absolute anchors on the remote machine, or
>the proxy will try to find them locally, and fail. Right now, we don't
>rewrite the pages passing through the proxy for http.
>Does output from mod_status show the child process stuck in the write
>state? If so, this problem should be fixed in 1.2b7, just released.
>Thanks for using Apache.
> wrote:
>> Submitter:
>> Operating system: Solaris 2.x, version: 2.4
>> Version of Apache Used: 1.2b6
>> Extra Modules used: proxy_module
>> URL exhibiting problem:
>> Symptoms:
>> --
>> Our server is configured to provide access to
>> internal pages as well as proxy requests to
>> our company wide firewall proxy.  The server
>> accepts requests from both port 80 and 8080.
>> All other options are default.
>> Our configuration file in the proxy section is as
>> follows:
>> ProxyRequests On
>> ProxyRemote ftp
>> ProxyRemote http
>> ProxyRemote gopher
>> ProxyRemote wais
>> The initial problem observed was that local pages
>> wouldn't load at random or individual images on a
>> page wouldn't load.  This cleared up with
>> a restart of the server for varying amounts
>> of time (minutes to hours).
>> With testing the problem appears to be that a
>> given child process stops handling local requests
>> after it handles a proxy request.  The process
>> will receive local requests and log them properly
>> but it returns nothing to the client.  The child
>> process will continue to handle additional proxy
>> requests properly.
>> Lynx, Netscape and direct telnet to port 80 have
>> all been used as the client and all have had the
>> problem.
>> --
>> Backtrace:
>> --
>> --
>Chuck Murcko
>The Topsail Group, West Chester PA USA

----- End of forwarded message from Siler, Vernon R. -----

Chuck Murcko            The Topsail Group             West Chester PA USA

View raw message