Bash script to monitor server load, mail from command line

[agntsgotnosecret]

Hello,

I am attempting to write a very basic script that monitors a server's load and automatically emails an administrator upon the load reaching a certain threshold...

So far I have:

#load=`uptime | awk -F ' ' {'print $10'} | cut -d ',' -f1 | awk -F '.' {'print $1'}`
if [ $load -gt 4 ];
then
mail -s "WARNING: Server Load Threshold Reached" $EMAILADDRESS
fi
exit 0

I'm running into a few problems, however...

1.) `uptime | awk -F ' ' {'print $10'} | cut -d ',' -f1 | awk -F '.' {'print $1'}` -- the output from this command results in a decimal figure, so when that value is parsed and placed in an if statement, the value is not seen as a number.

load-monitor.sh: line 9: [: 4.96: integer expression expected

How can I allow a number like "4.56" to be seen as an actual number within the if statement and be compared to 4, for instance?

2.) I am running a cPanel server on CentOS 5.4 -- when I run mail -s "SUBJECT" $EMAILADDRESS the command just hangs, and stracing the process shows it stuck on a read syscall.

3.) If I wanted to write this script in PHP, I have one primary confusion -- how can I mimic the functionality of obtaining the output from uptime via awk, etc. so that I can determine what the server load is at a given time? Which PHP function(s) would assist in that regard?

[Irithori]

Is writing the script yourself mandatory?

If not, there are enough monitoring solution for these kind of problems.
One that is quite simple to setup is Munin - Trac

You can do email alerting
HowToContact - Munin - Trac

and you have these nice rrd graphs, that show load/etc over time and therefore helping in getting a better overview of the system.
Munin :: ping.uio.no :: camilla.ping.uio.no


(for example: "Why is there a IO spike every night at 23:00" Ah, backup starts)

[agntsgotnosecret]

Hello,

Well, to me it's necessary simply for the educational aspect.

But thanks for the information.

[nmset]

Munin is a great tool, unfortunately it only produces graphs and you cannot store the actual data. Moreover, you cannot tweak the 5 min sampling interval.

For agntsgotnosecret problem, you can compare the load value as follows :

res=$(bc <<EOF
$load > 4
EOF
)

The comparison result will be 0 or 1, self-explicitly.

[agntsgotnosecret]

<code>res=$(bc <<EOF
$load > 4
EOF
)
</code>

Pardon my ignorance, but where exactly would you insert code into my script?

[nmset]

load=`uptime | awk -F ' ' {'print $10'} | cut -d ',' -f1 | awk -F '.' {'print $1'}`
res=$(bc <<EOF
$load > 4
EOF
)
if [ "$res" = "1" ];
then
mail -s "WARNING: Server Load Threshold Reached" $EMAILADDRESS <<EOF
EOF
fi
exit 0

Here Documents