|
On Tue 03 May 2005 21:40:42 BST , Joe Stump <joe@xxxxxxxxxxxx> wrote:
What I'd like is something that will allow me to do the following (for now):
1.) Monitor load on the servers (connections and actual load)
Configure snmpd on each machine to do just that, then poll it remotely
with the collector of your choice - I'd recommend Cacti for its'
"touch-free" nature after being installed, being entirely web-based.
2.) Restart Apache after I've made a change
On each realserver, run a script something like this from cron every <period>:
#!/bin/sh
#
# check httpd.conf for changes and gracefully restart Apache
CONF="/path/to/nfs/mount/httpd.conf"
OLDCONF="/path/to/tmp/dir/on/realserver"
HOSTNAME=`hostname`
DIFF=`diff -u $OLDCONF $CONF 2>&1>/dev/null`
RC=$?
if [ $RC -eq 1 ]
then
cp $CONF $OLDCONF
echo "Restarting Apache process on $HOSTNAME"
apachectl graceful
elsif [ $RC -eq 2 ]
echo "Problems found doing file diff on $HOSTNAME, bailing out..."
exit
fi
It means that you move the intelligence out to your Apache servers, so
if one of them is down you don't have a central control agent hanging
up trying to reach a problem server. You can extend that logic further
to monitor load also, especially if your servers are doing one job:
"If my 5 min load average exceeds THRESHOLD, stop Apache and flag to
the admin"
Clearly if the load average is sky high you could hit a problem running
this check, but then if it's that high you probably couldn't login from
your management station anyway...
HTH
Graeme
|
