Unix Job not failing as I would expect even though I have "Exit 1" in my script

[John O'Mullane]

Hi,
 I have a Unix shell script which has completed successfully but I expected it to fail.

I have an Exit 1 in my code 

echo "Checking DNFS....."

   count=`sqlplus -s '/ as sysdba'<<EOF

   set pagesize 0 feedback off verify off heading off echo off

   select count(*)  from v\$DNFS_stats where nfs_read >0 or nfs_write >0;

   EOF`

   if [[ "${count}" = '0' ]]; then

      echo "DB NOT running on DNFS"

      exit 1;

   else

      echo "DB running on DNFS"

   fi

   

   #jdkv=`cat /tmp/$1_jdkver.log`

   jdkcnt=`cat /tmp/$1_jdkver.log|wc -l`

   jdkver=`expr $jdkcnt`

   if [ $jdkver -eq 1 ]; then

      echo "Java working query passed"

   else

      echo "Java working query failed, Please check JAVA"

      exit 1;

   fi

When I run the Job, it's getting to this place in the code and exiting, but not failing.


https://us.v-cdn.net/5019921/uploads/editor/8p/awos96ulnqn5.png" width="808">

I have done a little test myself with a simple unix Job I created with only this code

https://us.v-cdn.net/5019921/uploads/editor/7a/7mm4qhcic2ia.png" width="482">

When I called from Automic, it behaved as I expected, job failed.

https://us.v-cdn.net/5019921/uploads/editor/sd/mm666odreb7o.png" width="520">


Has anyone seen this behavior before?

John.

[Legacy User]

JohnO'Mullane 
If you create a shell script for the first script and execute it from the command line, what is the return code?
For example:

./youscript

Then, execute:
echo $?

[Anon Anon]

just a silly idea:

use EXIT 1; instead of exit 1;

stolen here:
https://unix.stackexchange.com/questions/240723/exit-trap-in-dash-vs-ksh-and-bash

[John O'Mullane]

Thanks bobby_tamayo_automic FrankMuffke

Will try both those suggestions. 
Might be a few days before I get to test as it's not simple to reproduce the scenario in question.

[Matthias Schelp]

Hello JohnO'Mullane

You may have found proof for this:
 
https://community.automic.com/discussion/comment/33567#Comment_33567

Regards, Matthias

[John O'Mullane]

Thanks MatthiasSchelp

Carsten_Schmitz_7883
I have read your article below a couple of time but still not 100% clear.

https://community.automic.com/discussion/comment/33567#Comment_33567

Are you saying that the behavior I am seeing is intended functionality?

[Matthias Schelp]

Well, good luck with that. I'm pretty certain he won't answer you the next few days :smile:.

To make it short. What he means is that it could happen that Automic does not fetch the return code properly when you use "exit" or "exit 1".

Matthias

[Anon Anon]

provocative question:

why does it work _always_ in unix jobs that use bash?

... as long as I could observe this....

[Matthias Schelp]

:p Thats why I said we may have found proof for this.

[Anon Anon]

ahhhhh [shining bulb emoticon]
sry, fingers were faster than brain....

[Carsten Schmitz]

Well, good luck with that. I'm pretty certain he won't answer you the next few days :smile:.

Sorry, was gone there for a while, saving the galaxy. Or on christmas holiday. One of these.

JohnO'Mullane the post is intended to point out that the "Job messenger" doesn't get triggered when you use "exit" in your shell script. Our Automic consultant said to this: "don't use "exit" in your scripts". I kinda take that as the official Automic statement, but I find it rather limiting.

From my tests (and Wolfgangs), using "exit" does not prevent the parsing of exit codes. Something that exits with "exit 1" should show up as ENDED_NOT_OK in Automic, and it did many times for me, in practice. So I don't think that what I describe in the article is to blame.

But it's not entirely inconceivable that, under some circumstances, the lack of calls to the "job messenger" might cause this, e.g. when using different shells, so MatthiasSchelp is abolutely right in pointing out the possibility.

If you can still recreate the problem, and want to get to the bottom of it, you could use the trap approach I outlined further down in the post you linked. That way, you can "exit 1" and call the job messenger, so you could rule out whether the lack of a job messenger call is the culprit of the strange behaviour you saw.

Hope this helps.

[John O'Mullane]

Thanks Carsten_Schmitz_7883

A lot of our scripts (well, the scripts are not owned by us but are owned by the applications folk who use Automic) do have Exit 1's and work as expected. I haven't seen Automic not handle these Exit codes before.

Maybe this is a once off.

It won't be easy to reproduce I would imagine but will try and updates with findings.