In addition I see the following when doing a graceful restart of httpd:
[Mon Jun 26 14:17:14 2006] [notice] Graceful restart requested, doing
restart
[Mon Jun 26 14:17:15 2006] [error] (9)Bad file descriptor:
apr_socket_accept: (client socket)
[Mon Jun 26 14:17:59 2006] [notice] Digest: generating secret for digest
authentication ...
[Mon Jun 26 14:17:59 2006] [notice] Digest: done
I find very little on what the "Bad file descriptor: apr_socket_accept"
error means.
I am using Apache/2.2.0 on Linux 2.6.16-1.2133_FC5smp
Jim Albert wrote:
I'm hoping someone on the development list can comment on this.
I've noticed these in httpd access log starting with Apache2.2:
::1 - - [23/Jun/2006:04:03:04 -0700] "GET / HTTP/1.0" 200 117 "-"
"Apache/2.2.0 (Fedora) (internal dummy connection)"
I see a large number of those internal dummy connection requests during
an apache graceful restart (SIGUSR1) and at the same time the cpu load
on the Apache2.2 server maxes out at nearly 100%. I do not see this cpu
load during a graceful restart on apache 2.0 httpd servers.
I'm wondering if anyone else sees this high cpu usage with a graceful
restart of Apache2.2.
My home page (GET /) is dynamically generated (although running under
mod_perl) but I use the following mod_rewrite rule to make / a very
uncostly http request when the HTTP_USER_AGENT is "internal dummy
request". I'm not sure it is all those requests of / that is causing the
cpu load, but I created the RewriteRule, just in case.
RewriteEngine on
RewriteCond %{HTTP_USER_AGENT} ^.*internal\ dummy\ connection.*$ [NC]
RewriteRule ^/$ /small_static_page.html [L]
Can anyone comment or confirm a similar experience with high cpu usage
during a graceful restart of Apache2.2?
My OS is Linux 2.6.16-1.2133_FC5smp
--
Jim Albert
