[SOLVED] Alerting for specified condition

ankitpandey · 11-05-2012, 09:20 AM

Hello All,

I have created one script to send me alert mail when the run time of "particular process" goes beyond defined limit. How can i automate this, i mean i want to run this in such a way that it should send me the alert everytime the condition is met. Please let me know how will script keep executing so that as many times the condition is met it will send me alert mail.

Regards,
Ankit

theNbomr · 11-05-2012, 11:49 AM

Are you sure about that? You may need to reconsider your requirements.

You are saying that when a value exceeds some defined threshold, a mail message should be sent. Following that event, the value continues to exceed the threshold, so a mail must again be sent. How quickly do the second and all subsequent e-mails need to be sent? Or should the algorithm be different?

You can check the alarm status on a periodic basis (1-minute granularity) using cron to launch your process monitor. It doesn't seem to make sense to send an e-mail every time the alarm condition is true.

--- rod.

ankitpandey · 11-05-2012, 12:08 PM

Thanks for the response.

After first alert mail is sent, the second one should follow after 10 mins. I don't have permission to use cron job, can i do this without using cron job ?

Regards,
Ankit

theNbomr · 11-05-2012, 01:22 PM

You should be able to create a script that sleeps for 10 minutes (600 seconds), and then performs the test, sending an e-mail as necessary.

Code:

#! /bin/bash

while [ 1 ]; do

   echo "Checking alarm value..."
   #
   #  Run your test here, setting global variables, 
   #  or whatever works for your test

   if [ alarmValue -gt alarmLimit ]; then
      #
      #   Send e-mail
      echo "Sending e-mail to ankitpandey"
   fi

   #
   #   Wait ten minutes for next test
   sleep 600

done

--- rod.

SaintDanBert · 11-05-2012, 01:32 PM

Quote:

Originally Posted by ankitpandey

Thanks for the response.

After first alert mail is sent, the second one should follow after 10 mins. I don't have permission to use cron job, can i do this without using cron job ?

Regards,
Ankit

Your script can use a pattern similar to the following:

do {
if( alarm_exists() ) {
complain( "date", "time", "message" );
sleep( NAP_TIME );
until ( stopRunning == TRUE );

This can run at the user level so long as 'alarm_exists()' can learn what is needed
from your process in question. The longer the value of NAP_TIME, the smaller the
system load your script presents and the coarser the granularity of your complaints.
Any of the popular scripting languages will implement similar code.

The above approach does polling for the error condition. Depending on your process,
you might be able to enable and configure an "application event" such that your
script waits for that event and responds when it happens. I think this involves 'dbus'
but I don't know. Such an event catcher might need to field multiple events and complain every Nth occurrence. NOTE -- I've seen this done, but don't know the current implementations.

Let us know what you sort out,
~~~ 0;-Dan

SaintDanBert · 11-05-2012, 01:40 PM

Quote:

Originally Posted by ankitpandey

Thanks for the response.

After first alert mail is sent, the second one should follow after 10 mins. ...

Are you wanting the email to nag you until you respond somehow?
In that case, your process creates the error condition.
Your watchdog script notices and makes its complaint announcement.
After a complaint, the watchdog then looks for some acknowledgement.
After SOME_TIME, the watchdog repeats its complaint until it
gets the desired acknowledgement.

Logically, ERROR_EXISTS is a separate state from COMPLAINT_SENT which is
separate from COMPLAINT_ACKNOWLEDGED.

Think about your automobile.
Most ding or buzz with the keys in the dash and the door open.
You want the door open and you know where the keys are but that
blankety-blank-bleeping dinger just keeps making noise. Wouldn't you
like some way to say, "Thanks, I know where the keys are." to stop
the nagger? Would it be better if once acknowledged, you'd get a
fresh complaint after SOME_TIME if the keys remain in the dash?

Good luck. Keep us posted,
~~~ 0;-Dan