Restarting Apache hangs SSH session

Question

I have a script which will do a lot of prep work (generate config files, minimize and gzip Javascript and CSS files, etc) and then restart apache which is installed in a non-standard location). But after the script is run, when I try to exit out of my ssh session it just hangs indefinitely.

I've read the recommendation that I should redirect the IO for the actual call to the httpd binary with something like this:

sudo /path/to/httpd -f /path/to/config < /dev/null >& /dev/null

And this works and my ssh session no longer hangs after the script is done. But the problem is that now I've lost any errors that are spit out to STDERR if something goes wrong when it's started. This just seems really odd to me because I have another system with a similar (but not exact) setup that doesn't hang. And from reading the recommendation that suggests this workaround it says that daemons (like apache) should already be taking care of this. So what am I missing? Any pointers on how I can track this down or how I can fix it?

Answer 1

Firstly, you seem to have the ampersand in the wrong place. Instead of

/path/to/httpd -f /path/to/config < /dev/null >& /dev/null

I would expect

/path/to/httpd -f /path/to/config < /dev/null > /dev/null &

EDIT - actually, command >& file is a short form of command > file 2>&1. I've never used the short form, so I misunderstood the intent of that construct.

But I wouldn't do this. Redirecting STDIN and STDOUt to /dev/null is wrong. Omitting STDERR is also wrong. Doing this to program which puts itself into daemon mode is wrong. If you really had to do this I'd do something like

nohup /path/to/httpd -f /path/to/config >/tmp/httpd.out 2>/tmp/httpd.err &

But wait, Apache can be restarted by sending it signals using the apachectl script. You can signal Apache to gracefully reload it's configuration without dropping existing connections or you can signal it to restart afresh. If all you are doing is changing it's config, just do a graceful reload. I'd also use apachectl to do a configtest.

If Apache is in a non-standard location, you should still be able to use apachectl.

If you can't use apachectl you can send SIGUSR1 (reload) or SIGHUP (restart) signals to the main httpd process.

If apache isn't running, the best way to start it is by using the /etc/init.d/apache start script

If apachectl and the startup scripts don't work because you've relocated programs, I'd edit those scripts, there are variables that say which directories the executables are in.