You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by "Roy T. Fielding" <fi...@kiwi.ics.uci.edu> on 1997/11/01 02:11:37 UTC

night of the dead Apache

Hmmmm, something is wrong with 1.3b3-dev.  When reading my normal
test file (a page with about 120 small inline images), it seems to
pause every 10-20 requests for about a half-second.  Afterwords,
if I try to request /server-status, it *sometimes* just sits there
spinning -- all the server children and parent are sleeping, yet
Netscape (3.01/X) is waiting for a response.  I then did the same
request via telnet (in parallel) and am getting no response from
the server at all.  It connects, but nothing happens, and nothing
shows up in the error_log with full logging enabled.

5 minutes later, both requests are still dead.

MinSpareServers 2
MaxSpareServers 4
StartServers 4
MaxClients 150
MaxRequestsPerChild 60

no proxy, no vhosts, basic Apache+mod_info+mod_status

It recovers immediately after a kill -USR1 or kill -HUP (I tested both).
It is repeatable, but I'm not sure what triggers it.  It may be related
to keep-alive, since I have not been able to repeat it when there are no
children in keep-alive state.

I'll try repeating it with httpd -X, but I doubt that will work.

....Roy

Re: night of the dead Apache

Posted by Dirk-Willem van Gulik <Di...@jrc.it>.
Roy,

Could you put up this page (as a tar.gz or something) and
its gifs at hyperreal or somewhere else. That would be
usefull. You seem to be more able to reproduce. I can
see it sitting in read_with_errors or there about;
but cannot catch it with -X.

Dw.



Re: night of the dead Apache

Posted by Dean Gaudet <dg...@arctic.org>.

On Fri, 31 Oct 1997, Roy T. Fielding wrote:

> Hmmmm, something is wrong with 1.3b3-dev.  When reading my normal
> test file (a page with about 120 small inline images), it seems to
> pause every 10-20 requests for about a half-second.  Afterwords,
> if I try to request /server-status, it *sometimes* just sits there
> spinning -- all the server children and parent are sleeping, yet
> Netscape (3.01/X) is waiting for a response.  I then did the same
> request via telnet (in parallel) and am getting no response from
> the server at all.  It connects, but nothing happens, and nothing
> shows up in the error_log with full logging enabled.

Solaris lets you connect before its passed things on to the application. 
Can you truss the parent to find out if it's off in la la land?  It sounds
like it should be spawning a new child here. 

Dean