Script to get fail when there is an error in log file

Hi

I am not very good at shell scripting.and need you help.
Requirement is to write a shell script which which should get fail if there is any type of error/exception found in the log and the job should get failed without proceeding further.

1. We don't answer homework questions, but happy to assist if you show us your efforts.
2. Your requirement is pretty muddled. Please rephrase it better with examples.
3. Have a look here and here for questionnaire guidelines

With this information, all I could do is writing pseudo code. I assume that the log file is a text file. My solution:

if grep -q ERROR-TEXT LOGFILE
then
    FAIL-JOB
fi

To write better code, I would need to know:

What is "any type of error/exception"? For example: Does the word "ERROR" occur in the log file? But "ERROR" might be part of a non-error message. Is there a certain code that identifies an error or exception? Do log messages have a precise structure with a specific error field? etc.

Which job should get failed?

How can this job be failed - sending it a signal? Writing something to the job's input? Executing a certain program?

Originally Posted by berndbausch With this information, all I could do is writing pseudo code. I assume that the log file is a text file. My solution: if grep -q ERROR-TEXT LOGFILE then FAIL-JOB fi To write better code, I would need to know: What is "any type of error/exception"? For example: Does the word "ERROR" occur in the log file? But "ERROR" might be part of a non-error message. Is there a certain code that identifies an error or exception? Do log messages have a precise structure with a specific error field? etc. Which job should get failed? How can this job be failed - sending it a signal? Writing something to the job's input? Executing a certain program?

===

The error can be any oracle error/java exception error.
So,Ii want to grep error, ora,fatal,exception etc in the log file.
there is a script runs via a job scheduler and I want whenever there is any such error the job should get fail.
The job is running in the server with respective process id.
So, in case of any error identified in the log file , job should get fail immediately.

If I've quite understood you, you would like to stop the job according to what the log is displaying?
Isn't the job aware of what it asks the log to store? Because if it is, maybe the job could stop itself while generating the corresponding log?
Anyway, as stated by Honest Abe in post#2, you should show us some efforts...but so far you have just provided us with requirements.

Amended script, still lots of gaps for you to fill:

if grep -q -e error -e ora -e fatal -e exception $LOGFILE
then
    kill $JOB_PID
fi

You will have to correct the grep search. The way I wrote it, you will have plenty of false positives. You need to analyze the error messages and craft water-tight regular expressions that find those errors, and only those errors.

Another gap is the job PID. That should be easy, but you need to find out how to stop it gracefully. I just assume that a termination signal does the job.

Write your script, test it, and if you get stuck, ask again.

$ cat xyz.ksh
#!/bin/ksh

V_LOG = /cs/mantas/fccm801/database/db_tools/logs/äbc.log
V_ERR = egrep -wi 'Exception|Error|ORA-*|FATAL|FAIL' $V_LOG

STATUS=$?

case $STATUS in
1)
echo "Exception raised: of type Exception"
;;
2)
echo "Exception raised: of type SQLException"
;;
esac

return $STATUS

if ($V_ERR -gt 0)
then
echo "Check detailed error in the log file"
exit -1;
else
echo "Job is successfully completed"

It appears as though you're guessing. How does any value get assigned to $STATUS? I don't think we understand what you mean by 'fail(ed)'. Your script doesn't show anything related to killing a process. It appears it only gives a message. I'm not sure if the -gt 0 will get you anywhere, because the variable is nonnumeric.

This might get you going a bit more: http://www.well.ox.ac.uk/~johnb/comp/unix/ksh.html