From nobody@hyperreal.com Wed Mar 12 20:46:02 1997 Received: by taz.hyperreal.com (8.8.4/V2.0) id UAA04022; Wed, 12 Mar 1997 20:46:02 -0800 (PST) Message-Id: <199703130446.UAA04022@taz.hyperreal.com> Date: Wed, 12 Mar 1997 20:46:02 -0800 (PST) From: Bradley White Reply-To: brad@lightspeed.net To: apbugs@hyperreal.com Subject: child process dump core X-Send-Pr-Version: 3.2 >Number: 231 >Category: os-bsdi >Synopsis: child process dump core >Confidential: no >Severity: critical >Priority: medium >Responsible: apache >State: closed >Class: sw-bug >Submitter-Id: apache >Arrival-Date: Wed Mar 12 20:50:00 1997 >Last-Modified: Sat Jun 28 15:38:46 PDT 1997 >Originator: brad@lightspeed.net >Organization: >Release: 1.2b7 >Environment: BSDI 2.1 Fully Patched and BSDI 3.0 dist. gcc is the compiler >Description: When configuring Apache for more than 125 virtual hosts that use one single ip address http1.1 style, the child process start dumping core. It's interesting that the parent process does not dump core. The syptoms of this are are web pages that should be working, coming up as either connection reset by peer or document contains no data. A check of the error_log while this is going on shows the child process giving out a segment violation. >How-To-Repeat: Configure a bsd/os server with 126 or more virtual web servers http1.1 style. We are unable to test with any other OS at this time. >Fix: N >Audit-Trail: State-Changed-From-To: open-feedback State-Changed-By: dgaudet State-Changed-When: Tue Mar 25 12:57:10 PST 1997 State-Changed-Why: While this is most definately a problem with too few file descriptors apache should not core dump because of it. If possible could you provide us with a stack trace from gdb on the core dump? To avoid the core dump you'll need to either increase the number of descriptors available (see the limit command) or run multiple apaches and use Listen to spread the vhosts amongst them. State-Changed-From-To: feedback-closed State-Changed-By: dgaudet State-Changed-When: Sat Jun 28 15:38:46 PDT 1997 State-Changed-Why: A workaround and documentation has been committed to 1.2.1-dev and 1.3-dev. Thanks Dean >Unformatted: