在 Robot Framework 中运行一段时间的测试用例

Question

I wanted to run robot test for a duration of time, say for 1hr. No matter if the execution of all test cases in a test suite is completed. It should repeat the test cases until the given time reached.

I tried to use --prerunmodifier and tried to write my own module, I used robot.api module robot.running.context , and override the present method end_suite() . But not successful! :(

Answer 1

As per the Robot framework user guide the robot framework test suite will be executed for a maximum time of 120 minutes ie 2hrs. Not we can overwrite this timeout by explicitly mentioning the test execution time in the test files Settings table as below

***Setting***
Test Timeout  60 minutes

Further you are most welcome to set a test cases specific timeout using [Timeout] option as below

***Test Cases***
Sample Test Case
  [Timeout]  5 minutes
  Do some testing
  Validate some results

Feel free to try and ask further questions.

Answer 2

Try with ' Repeat Keyword ' keyword. It takes as argument for how long it should repeat given keyword. But in this case all your test cases should go to one keyword. Use 'Run Keyword And Ignore Error' inside of it so you ignore errors.

Eg:

Repeat Keyword   2h    Keyword With All Test Cases

Second option would be writing a Listener - has similar functionality as prerun modifier but is executed during tests not before.

Answer 3

I would approach this with an outside runner script using two threads. Sample algorithm with no implementation:

Start a time counter (just the current epoch time); the first (the main) thread is a controlling one. It sleeps for 1s, and on run checks and branches:

• the counter has not reached the threshold (eg current epoch - starting < limit) - check is the other thread running (has not finished execution):
  -- yes: pass

  -- no: that "other" thread is the execution of the robot suites. It is not running - spawn a thread that will run the suites; before that - copy the logs from the previous execution - if you care about them (in general I guess you would, to analyze the run's result; except if you're doing longevity perf test, and want to monitor just the SUT). Thus you will re-start the suites after they finished.

• the counter has exceeded the threshold - stop the run in the other thread. For that the INT or TERM signal - this will allow the framework to create the log files. If you use KILL, that'll abruptly end the current execution - if you don't care about the logs that's an option. But you risk leaving your SUT in polluted state - the case that was running may have changed its configuration, and will not have the chance to clear up after itself.

I would advise on starting the execution to specify not to generate log and report html files from the intermediate runs (the --log none --report none CLI options) - their generation will take execution time, during which your cases will not be "spinning". After the run is done you can generate them from the intermediate output.xml files you've copied before.

If you are keen on this approach, and want to use python for the controlling script/system, I'd suggest looking at the source of pabot - it's that (threaded control and runs, but no timebound, your ask) on steroids.