cronjob questions

grob115 · 07-12-2006, 01:18 AM

Hello,

I have a forum software on my server that has been having trouble accessing the mySQL database, which is hosted on a different server. The problems can be either too many connections or lost connection. In an effort to restart Apache on the web server (this process somehow reset the connection to the mySQL server), the host has setup some cronjobs to do that. The following is a list of the user cronjob (apparently the "service httpd restart" line has not been working and so they added (cd /root; ./newapachestart.sh 2>&1) also):

Code:

[root@vps ~]# crontab -l
#*/10 * * * * /root/chkload.sh
#59 */2 * * * /root/sendcheck.sh
# following is not working in case of httpd already in stop stat !! yes.. strange..
#*/15 * * * * service httpd restart

*/5 * * * * (cd /root; ./newapachestart.sh 2>&1)

and ./newapachestart.sh contains:

Code:

#!/bin/bash

CNT=`ps ax | grep -v grep | grep -c httpd`

if [ $CNT -gt 10 ]

then
echo $CNT;
echo 'Restarting HTTPD...'

killall httpd
killall httpd
killall httpd
/etc/init.d/httpd restart

fi

if [ $CNT -le 0 ]
then

/etc/init.d/httpd restart

fi

Can someone please explain to me the following:
1) What exactly does (cd /root; ./newapachestart.sh 2>&1) mean? I don't quite follow why we have to change directory to root, and what is this "2>&1" business.
2) What is going on in this ./newapachestart.sh file? I really have no idea of what it's doing.
3) Is there a way to setup a .ps file to test the connection to mySQL and only restart if the connection is lost? (like maybe pinging the mySQL server to test the connection)

chrism01 · 07-12-2006, 01:31 AM

1. The cmd ./newapachestart.sh means run script found in 'current' dir (ie './' bit), so they change to /root where the script is (should be).
"2>&1" means copy stderr to stdout
2. script counts num of httpd processes, if cnt >10 logs a msg, kills them all (several times!), then does a restart.
If cnt <= 0, do a restart.
When I did my Apache admin course (1.3.x) at RedHat, they said that 'restart' could be a bit iffy, so always use 'stop' then 'start' instead. You should prob either wait a bit eg 5 secs, or check again eg in a loop, between stop and start; ie ensure Apache really has stopped before starting it.
iirc, theres a .pid file with the main process' pid in it.
3. .ps is usually a postscript file, so you've lost me there, but a ping or logon/off prog is certainly feasible.

grob115 · 07-12-2006, 02:08 AM

Hello,

Ah I see. Thanks. It makes sense now. Just a few further questions:
1) Why the need to kill the process when Apache is going to restart anyway? Wouldn't restarting Apache kill the httpd process?
2) Why kill several times? Isn't "killall httpd" kills all already?
3) How's "/etc/init.d/httpd restart" different from "service httpd restart"?
4) Sorry made a mistake about ps file. I meant to say a sh file. :P Would you mind illustrating how to do this "ping or logon/off prog" process in a .ps file that I can add to the crontab? All that I want to do is to check the mySQL connection every once in a while with cron and if it's lost, just log off/on.