Ability to Stop Locust Client from within the test script

[uprichard]

We have tests that for each client need to execute some POST's and then perform X amount of GET's and then perform some more PUTs before stopping. So this is not the normal feature request for adding back the 'stop after number of requests"

Currently when running in standalone mode I have a task that has a loop in it that executes the GET's X times. It then performs the final set of POST's and does a 'raise StopLocust' that will correctly stop the client

The issue is that this will not work in a master / slave setup as the 'StopLocust' will stop the client(s) correctly but the slave process will continue running until the RUN_TIME is reached.

So looking for a way that I could stop a given client that works in both standalone and master/client. This approach could also be used by others who need to stop after X requests, they could keep a counter on each call and when hit their value, stop the client.

Thanks for the great tool

[TanmayNakhate]

I support this as well. We are also trying to implement same behavior.
Although I understand that to handle this behavior, I will have call MasterLocustRunner.quit() but we have to provide certain parameter to its object which is difficult as of now. I am trying to find out if I am able to solve it but I am not yet successful.
My stack overflow question is here (below)... but there are very few people who view / answer .. :(

StackOverFlow Question

[cyberw]

Hmm... I'm not sure there is a way to do this atm. @heyman do you know a way to do this?

[TanmayNakhate]

Hmm... I'm not sure there is a way to do this atm. @heyman do you know a way to do this?

I attempted to print init params for Locust class, which takes 2 params - locust_classes, options.
If we can get the same object used at the time of initialization of first locust object, it may be possible. Although I attempted to create Locust class object manually, I am not succesful.

If it helps , this is the output of print statement of locust init :
locust clasess : [<class 'Locust_ncso_multiuser_task.NcsoLoad'>] #NcsoLoad is my class while Locust_ncso_multiuser_task is test name.

options: Namespace(csvfilebase=None, exit_code_on_error=1, expect_slaves=1, hatch_rate=1, heartbeat_interval=1, heartbeat_liveness=3, host=None, list_commands=False, locust_classes=[], locustfile='testsuite/host/Locust_ncso_ics_multiuser_task.py', logfile=None, loglevel='INFO', master=False, master_bind_host='*', master_bind_port=5557, master_host='127.0.0.1', master_port=5557, no_reset_stats=False, no_web=False, num_clients=1, only_summary=False, port=8089, print_stats=False, reset_stats=False, run_time=None, show_task_ratio=False, show_task_ratio_json=False, skip_log_setup=False, slave=True, stop_timeout=None, web_host='')

[TanmayNakhate]

Hmm... I'm not sure there is a way to do this atm. @heyman do you know a way to do this?

I attempted to print init params for Locust class, which takes 2 params - locust_classes, options.
If we can get the same object used at the time of initialization of first locust object, it may be possible. Although I attempted to create Locust class object manually, I am not succesful.

If it helps , this is the output of print statement of locust init :
locust clasess : [<class 'Locust_ncso_multiuser_task.NcsoLoad'>] #NcsoLoad is my class while Locust_ncso_multiuser_task is test name.

options: Namespace(csvfilebase=None, exit_code_on_error=1, expect_slaves=1, hatch_rate=1, heartbeat_interval=1, heartbeat_liveness=3, host=None, list_commands=False, locust_classes=[], locustfile='testsuite/host/Locust_ncso_ics_multiuser_task.py', logfile=None, loglevel='INFO', master=False, master_bind_host='*', master_bind_port=5557, master_host='127.0.0.1', master_port=5557, no_reset_stats=False, no_web=False, num_clients=1, only_summary=False, port=8089, print_stats=False, reset_stats=False, run_time=None, show_task_ratio=False, show_task_ratio_json=False, skip_log_setup=False, slave=True, stop_timeout=None, web_host='')

If we can stop locust runners in a better way, it would help as well.
Lets say i am using 10 clients and test_iterations=100. So I am assuming 1000 requests will be sent.
and after for loop runs for 100 times (9 other clients running same loop 100 times in back-end), I am explicitly calling "runners.locust_runner.quit()" , what happens is, it kills all other runners as well, thus breaking flow for all the other clients as well.

[cyberw]

Would it be possible/appropriate to stop locust if all Users have terminated in master-worker mode @heyman ?

[cyberw]

I had forgotten about this ticket but actually ended up writing a PR for this :) #1448

[cyberw]

Fixed by #1448

See https://docs.locust.io/en/latest/writing-a-locustfile.html#environment-attribute