[Ground-user] Grinder Headless Execution - automating test runs without the need for manual intervention

[Ouray V. <ou...@vi...>]

Hi All:

I am looking for some advice/suggestions for the following test/business
requirement.

As a test engineer, I need to be able to run Grinder in such a way that does
not require user intervention, in order to allow for unmanned testing.

Example:

I have 3 test runs that I need to execute.  I do not want to manually
monitor the tests from the Grinder Console.

Test methodology:

   1. Clean application logs
   2. Start-up application
   3. Test to ensure application is ready for load, avoid manual timing in
   scripts etc.
   4. Run a script (bash) that makes a call to grinder to start-up
   (headless - without the Grinder Console).
   5. Once the grinder run completes, harvest the required
   test artefacts and start back at bullet 1 to complete n runs.


Psydo code for wrapper script, in this example a shell script will do:

Description:  This script will run 3 tests, and follow my desired test
methodology for each test run.

TEST_NAME=Checkout
TEST_RUNS=1 2 3


for i in ${TEST_RUNS};
do
    # call application to startup
    # call to validate that the application is ready for testing
    # call to start resource monitoring on SUTs
    # call grinder agent to start
    # once grinder agent complets call
    # call to stop application
    # call to fetch logs
    # call to delete old logs
    # call to fetch resource logs on SUTs
done

Other options:
- pass in dynamic grinder properties upon calling grinder agent to allow for
changing the 1) script 2) duration 3) thread for example.

Call to script (see Script: section below):

./script.sh grinder.useConsole=false grinder.threads = ${THREAD_COUNT}
grinder.duration = ${TEST_DURATION}

Script:

<snippet>
for args in $@; do JAVA_ARGS = "${JAVA_ARGS} -D${args}"

$JAVA_HOME/bin/java -cp $CLASSPATH $JAVA_ARGS net.grinder.Grinder
$GRINDER_PROPERTIES
</snippet>

My guess:
- So far it seems that you can start an agent and tell it to start without
the Grinder Console, which intrinsically tells the agent to run a particular
script (as defined in the grinder.properties).  I have tested this with
Grinder Stone, and it works great.

My Question:
- Is there any know reason why this is not a good idea?  Has anyone seen/see
issues with this approach?
- Are there any better suggestions?

Any thoughts/comments are welcome.

Kind Rgds,

-- 
Ouray Viney
http://www.viney.ca