2

Recently, my site's visitor-ship increased sharply. My Apaches' always crash. I'm seeing lines like these in the httpd/error_log:

[Sun Apr 13 08:57:18 2014] [warn] child process 7792 still did not exit, sending a SIGTERM
[Sun Apr 13 08:57:20 2014] [error] child process 7925 still did not exit, sending a SIGKILL
...

I will be upgrading my server in the next month. Before that, I'm trying to write a Bash script to automatically restart Apache with a crond job, which checks every 5 minutes.

If curling some page does not return a 200 code, then fuser -k -n tcp 80 and /etc/init.d/httpd restart.

I'm not good at Bash scripting. I've found some working code which will only restart Apache. 

#!/bin/bash

curl -I http://www.mydomain.com/some-empty-page 2>/dev/null \
    | head -1 | grep -q " 200 OK"

if [ $? -eq 0 ]; then 
  echo ""
else 
  /etc/init.d/httpd restart
  echo "wrong $(date)" >> /home/myspace/restart_log.txt
fi

How can I modify this to kill all the jobs which use port 80 before restarting?

I'm intending to insert some working line like:

else fuser -k -n tcp 80
  /etc/init.d/httpd restart
  echo "wrong $(date)" >> /home/myspace/restart_log.txt
fi

NOTE: first else fuser -k -n tcp 80, then do /etc/init.d/httpd restart in the else case.

Improve this question
5
  • What happens when you try it? Commented Apr 13, 2014 at 20:04
  • There are more robust monitoring solutions (search Unix & Linux or Server Fault or the web for “Apache monitoring” or the like). Besides that, I don't understand what you're trying to do here. Just doing /etc/init.d/httpd restart will restart Apache, you don't need to kill anything else beforehand. Commented Apr 13, 2014 at 22:25
  • @glenn jackman, add else fuser -k -n tcp 80; /etc/init.d/httpd restart; the script can not run. I monitor the /home/myspace/restart_log.txt. Commented Apr 14, 2014 at 11:13
  • "the script can not run": what happens exactly? Commented Apr 14, 2014 at 11:24
  • i am monitor the /home/myspace/restart_log.txt,if no fuser -k -n tcp 80;, if the apache crash down, it will insert new line about wrong $(date), but if add fuser -k -n tcp 80; no new line insert into restart_log.txt and apache remains in a crash status. Commented Apr 14, 2014 at 11:27

1 Answer 1

Reset to default
3

I think you're misunderstanding the usage of port 80. Only one daemon is listening on port 80. It's then forwarding incoming requests to Apache worker daemons. I would simply run the restart script if you're curl command doesn't come back with a HTTP status 200.

The stop/start service script should already have provisions in it for dealing with Apache's that are up and not willing to come down.

If you're finding that even after it runs like so:

$ /etc/init.d/httpd stop
$ pgrep httpd
... returns process ID's...

Then you may want to turn your approach around a bit and do this instead:

$ /etc/init.d/httpd stop
$ pkill httpd
$ /etc/init.d/httpd start

But as I've said. You really should not have to do any of this. The service script /etc/init.d/httpd should be robust enough already to properly deal with these situations itself.

Modified script

I'd modify what you have in the following way, but otherwise it would be fine to use as is, IMO.

#!/bin/bash

response=$(curl --write-out %{http_code} --connect-timeout 5  \
    --silent --output /dev/null http://www.mydomain.com/some-empty-page)

if [ $response -eq 200 ]; then
  echo "All's well"
else
  /etc/init.d/httpd restart
  echo "wrong $(date)" >> /home/myspace/restart_log.txt
fi

If you want the killing feature, change the else clause:

else
  /etc/init.d/httpd stop
  pkill httpd
  /etc/init.d/httpd start
  echo "wrong $(date)" >> /home/myspace/restart_log.txt
fi

Depending on how severe a restart you want to perform, you can expand the pkill to give harsher kill signals (SIGKILL aka. -9) for example, but I would start with the former. If you experience hanging Apache's then add the else clause next. If that still leaves hanging Apache's then use SIGKILL's.

Improve this answer
3
  • thanks for your suggestion, so can you give me the completed working #!/bin/bash script? Commented Apr 14, 2014 at 7:13
  • thanks for your update anwser. but still some strange problem. the server still crashed. i tried code in ssh, /etc/init.d/httpd stop.............................., it always in waiting status, type pkill httpd before /etc/init.d/httpd restart, it will /etc/init.d/httpd restart.............................. always in waiting status. But type fuser -k -n tcp 80 then do /etc/init.d/httpd restart, the server can restart and worked. what is this problem? I met this problem also in mysql, fuser -k -n tcp 3306 can work, but pkill mysqld sometimes not working before restart. thx Commented Apr 15, 2014 at 9:17
  • Have you looked at these processes that are hanging? What do they look like via ps -eaf | grep httpd? Also I would take a look at one of them using lsof -p <pid> to see what resource they're hanging on AFTER you've given the /etc/init.d/httpd stop command. Perhaps you need to sleep 3 seconds to give them time to stop? Commented Apr 15, 2014 at 12:47

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.