Retry based on AssertionStep status
Hi there
Intro here, the actual problem is below the line. We are currently working on a retry-mechanism that re-runs a specific interval of testSteps upon failure.
In the given picture the test would transfer a file to a server, which extends an already existing list with new data. Because the processing system is outside of our scope we can only wait and guess when the server is done ingesting and processing the data. In the current project we wait for a static 2 Minutes (and it still takes longer sometimes).
The entire logic would be to simulate a testRunner and have the script run all steps from index 2 to 5, check the assertion step at index 5 and then continue the testCase. Upon failure it would run all steps from index 3 to 5 again.
The pure logic is rather basic and would look like this:
void ftpRetry(def testRunner, def context, int retries){
List retryCheckpoints = ['retryCheckpoint']
int currentIndex = context.getCurrentStepIndex()
( checkpointStep, checkpointIndex ) = findStepInTestCase(context, false, retryCheckpoints)
Boolean passCheckpoint = false
while(retries-- > 0 && !passCheckpoint){
runTestSteps(testRunner, context, currentIndex + 1, checkpointIndex)
if(checkpointStep.getAssertionStatus().toString() == 'PASS') {
passCheckpoint = true
}
}
assert passCheckpoint : "Out of retries, checkpoint-assertion failed every time!"
}
void runTestSteps(def testRunner, def context, int startIndex, int endIndex) {
while(startIndex <= endIndex) {
context.getTestCase().getTestStepAt(startIndex++).run(testRunner, context)
}
}
________________________________________________________________________________________
The script looks fine to me, but calling .getAssertionStatus() on the AssertionStep always returns "PASS". It appears as if the step itself runs, but the assertions within the step are not running because their status is always "UNKNOWN". This may have something to do with the AssertionStep itself, but we currently can not use internal assertions in REST-Steps themselves. We must put assertions in their own step!
I am currently crawling through the API to figure out why the assertions within the assertionStep don't want to run, but I'd like to see if someone can provide a solution.
I managed to solve this specific problem, the opened support request shall remain open because I still suspect a problem with the .run() function.
Instead of trying to emulate the testRunner from inside the groovy step we now "drop an anchor" before the retry-cycle, run into the script and check the checkpoint status. If the status is FAIL we tell the testRunner to go back to the anchor via testRunner.gotoStep() - if the status is PASS we just let the TestCase continue on. Here is how that looks:
The anchor contains the number of remaining retry attempts, which are subtracted with each iteration by the ftpRetry-Script.
And here is the adapted ftpRetry function from the example.
void ftpRetry(def testRunner, def context){ // variables to allow user to specify other names List<String> retryCheckpointNames = ['retryCheckpoint'] List<String> retryAnchorNames = ['retryAnchor'] int defaultRetryCount = 15 ( checkpointStep, checkpointIndex ) = findStepInTestCase(context, true, retryCheckpointNames) ( anchorStep, anchorIndex ) = findStepInTestCase(context, true, retryAnchorNames) if(checkpointStep.getAssertionStatus().toString() != 'PASS') { // checkpoint did not pass, get the anchor, set the testRunner to the index of the anchor and subtract a retry from the anchor, then let everything run until this script is called again int retries = anchorStep.getPropertyValue('retries').toInteger() if(retries-- > 0) { anchorStep.setPropertyValue('retries', retries.toString()) testRunner.gotoStep(anchorIndex + 1) } else { anchorStep.setPropertyValue('retries', defaultRetryCount.toString()) assert false : "Out of retries, checkpoint-assertion failed every time!" } } else { log.info("Checkpoint passed, resetting retries on anchor.") anchorStep.setPropertyValue('retries', defaultRetryCount.toString()) } }Should have reconsidered my approach to the problem rather than trying to force the script to emulate the testRunner, oh well.
