Apache Stops Without Evident Cause

Home » CentOS » Apache Stops Without Evident Cause
CentOS 7 Comments

OS = CentOS-6.4
Uptime = 4 days, 1:31, httpd-2.2.15
SELinux=enforcing audit2why -l -a = nil report

Since upgrading a public web service host from dedicated hardware running CentOS-5.x to a KVM virtual host running CentOS-6.3 I have experienced intermittent ( one every couple of months) halts by the Apache web service. There is never any trace left in any of the logs, there are no SELinux avc’s recorded, and the rest of the system remains up. It is just the httpd service that stops.

I have an automatic restart job in place so this is not a show stopper. But I am concerned that this behaviour is pointing to a fundamental problem somewhere.

Has anyone else experienced this? Are there any ideas as to what could be causing this?

7 thoughts on - Apache Stops Without Evident Cause

  • Define ‘stops’

    Are there any httpd processes still running?

    What are you doing with the server – php, proxy to tomcat, etc

    There’s very little information to go on in your post…

  • Attempts to access the web sites served by that service report Server not Found.

    Running ‘service httpd start’ starts the service.

    No.

    Straight-up static web sites. Nothing fancy or dynamic. No CGI scripts.

    You have what I have. There is simply no trace of the process by the time my periodic (every 15 minutes) sweep notices that the httpd process is gone and restarts it.

  • My first guess on mysterious process death on Linux is always the kernel out-of-memory killer. But I think that is logged somewhere.

  • James B. Byrne wrote:

    A possibly useful thought for something that almost always, for me, is useless: is crashkernel enabled? Any dumps?

    Or, for that matter, since I don’t know how it’s being run or who has what authorities, anything in sudo.log?

    Ooog – is selinux enforcing?

    mark

  • By default the OOM messages should go to the kernel ring buffer (so use dmesg to see). I think that they also go to syslog, so should also be in /var/log/messages.

    –keith

  • Keith Keller wrote:

    They do, indeed, go to messages. I’ve seen them *far* too frequently here
    (serious heavy-duty multi-threaded scientific computing, or R, will eat
    *everything*).

    mark

LEAVE A COMMENT