Re: daily apache-ssl reload is causing probs (FIXED)
Ok, never mind. Everything below is correct except for the solution.
no matter how I get that SIGUSR1 sent to apache-ssl, it still fails the
reload randomly - from the command line or from logrotate...
I guess I'll be trying apache+mod_ssl out of despiration.
On Tue, May 06, 2003 at 10:33:34AM -0600 or thereabouts, David Wilk wrote:
> Howdy folks,
> Well, I did some digging and found some answers. I'm posting my
> solution here in the event that others might find it useful.
> first of all, there's a bug in woody's logrotate package. Logrotate
> will (might?) issue the postrotate command from any logrotate config (be
> it daily, weekly or monthly) on a daily basis. So, if you only expect
> apache-ssl to get restarted (apache-sslctl restart or
> /etc/init.d/apache-ssl restart) monthly with your monthly logrotate
> config, think again. it's probably happening every night.
> Second, I figured out that although the debian apache-ssl script
> (/etc/init.d/apache-ssl) would cause apache-ssl to choke on a reload,
> using 'apache-sslctl graceful' which has the same effect of sending a
> SIGUSR1 to the apache-ssl parent process does not appear to have the
> same problem. So far I've gone through several weeks of daily
> apache-sslctl graceful's and one monthly logrotate without a single
> problem with apache-ssl. so, to be clear, all I did was replace
> /etc/init.d/apache-ssl reload with /usr/sbin/apache-sslctl graceful in
> the postrotate command in my logrotate config for apache-ssl.
> I haven't investigated exactly why this change works, but I know that it
> does, most likely. I won't really be sure that it's trouble free until
> it's been up for several months without incident, but after going down
> every couple days, I feel like a few weeks is a pretty good indication
> that the problem has been resolved.
> hope someone finds this useful.
> On Wed, Apr 16, 2003 at 11:56:45AM -0600 or thereabouts, David Wilk wrote:
> > Hello all,
> > I think I have found that an /etc/init.d/apache-ssl restart is the only
> > way to properly restart apache-ssl after a logrotation. However, I've
> > had apache-ssl die two days in a row, and the culprit appears to be some
> > process that is sending apache-ssl a SIGUSR1 (what apache-ssl reload or
> > httpsdctl graceful issues).
> > Here's the log:
> > [Mon Apr 14 03:00:18 2003] [notice] SIGUSR1 received. Doing graceful
> > restart
> > [Mon Apr 14 03:00:18 2003] /usr/lib/apache-ssl/gcache started
> > [Mon Apr 14 03:00:19 2003] [error] (2)No such file or directory:
> > mod_mime_magic: ca
> > n't read magic file /etc/apache-ssl/share/magic
> > [Mon Apr 14 03:00:19 2003] [notice] Apache/1.3.26 Ben-SSL/1.48 (Unix)
> > Debian GNU/Li
> > nux PHP/4.1.2 mod_perl/1.26 configured -- resuming normal operations
> > [Mon Apr 14 03:00:19 2003] [notice] suEXEC mechanism enabled (wrapper:
> > /usr/lib/apa
> > che-ssl/suexec)
> > [Mon Apr 14 03:00:19 2003] [notice] Accept mutex: sysvsem (Default:
> > sysvsem)
> > the problem is I don't know what could possibly be issueing this SIGUSR1
> > signal to apache-ssl every morning at the exact same time that cron runs
> > /etc/cron.daily. I've checked all my cron jobs and can't seem to find
> > the culprit.
> > if anyone has any ideas, I'd be grateful.
> > thanks,
> > Dave
> > --
> > *******************************
> > David Wilk
> > System Administrator
> > Community Internet Access, Inc.
> > email@example.com
> > --
> > To UNSUBSCRIBE, email to firstname.lastname@example.org
> > with a subject of "unsubscribe". Trouble? Contact email@example.com
> David Wilk
> System Administrator
> Community Internet Access, Inc.
> To UNSUBSCRIBE, email to firstname.lastname@example.org
> with a subject of "unsubscribe". Trouble? Contact email@example.com
Community Internet Access, Inc.