Re: stops listening to IP (Apache Users)

On Aug 12, 2014, at 12:57 PM, Jeff Trawick wrote:

On Tue, Aug 5, 2014 at 7:31 PM, Al <al@xxxxxxxxxxxxxxxxxxxxxx> wrote:

Hi,

On Aug 5, 2014, at 4:51 PM, Jeff Trawick wrote:

On Tue, Aug 5, 2014 at 3:27 PM, Al <al@xxxxxxxxxxxxxxxxxxxxxx> wrote:

Hi,

I am running apache 2.4. We are now having a problem with the server, it stops answering requests on one IP. It goes down about every 45 minutes. I have since switched one of the websites to another IP and all it did was transfer the problem to the other IP. I am now using both IPs for that website and I have disabled keep alive. It now needs to be restarted every 5 hours. Does anyone have any ideas as to what is going on?

operating system?
Listen directives?
any IPv6 interfaces?
which MPM?
anything in error log?
anything in system log related to the interface that no longer works with httpd?

I should have included this info. OS NetBSD 5.2, Listen 0.0.0.0:80 Listen 0.0.0.0:443, no IPv6, prefork, nothing in any of the logs that I can find.

APR had an issue with feature detection on FreeBSD 10 that led to "outage" symptoms, though I don't recall precisely how similar the symptoms were. Do you know if NetBSD 5.2 has the accept4 syscall? (doubtful) That issue for FreeBSD was fixed in apr-1.5.1.

With those Listen directives, httpd should be immune from different behavior on different interfaces, even in the presence of bugs related to socket state, polling, etc.

You could attach to children with gdb at the time of the outage and get backtraces so we can see if httpd is doing its best to handle new connections. If you can figure out how to trigger the issue, you could do so while running strace/truss/(similar tool) against the httpd child processes, which is trivial if you're using a threaded MPM and configure it to only have a single child.

Other than that, I dunno.