On Stop causes the task to continue

[masonCaminer]

Describe the bug

If I add a new table concurrently, I click Stop, and then my on_stop method will be executed. On_stop is deleting the table. At this moment, one user will start executing on_stop and the other user will continue to execute task

Expected behavior

Task is no longer executed when on_stop is expected

Actual behavior

The Task is still executed

Steps to reproduce

1. Task is a new table
2. On_stop removes the table one by one

Environment

• OS:
• Python version: 3.8
• Locust version: 1.2.3
• Locust command line that you ran: locust -f src/loadTest/11_add_core.py --users=20 --spawn-rate=20
• Locust file contents (anonymized if necessary):

[cyberw]

can you add your locust file?

[cyberw]

@masonCaminer I'm super interested in debugging this, as I've had similar issues in the past, but right now I'm unable to reproduce. Can you post your locust file? Also, what "tables" are you talking about?

[masonCaminer]

This is our project. I can't provide you with the process of pressure testing. What I can provide you is the process and script

[masonCaminer]

USER_CREDENTIALS = PerformanceTest.get_users()
PerformanceTest.init_locust_env()
setup_logging(loglevel="INFO")


class QuickstartUser(FastHttpUser):
    wait_time = between(1, 1)
    host = os.getenv('host')

    def on_start(self):
        if len(USER_CREDENTIALS) > 0:
            name, password = USER_CREDENTIALS.pop()
            self.login = Treelab.login(email=name, password=password)
            self.workspace = Treelab.get_workspaces(token=self.login.token)[0]
            self.core = self.workspace.add_core(core_name=random_str(), is_assert=False)

    @task
    def add_table(self):
        self.core.add_table(table_name=str(uuid.uuid4()), is_assert=False, locust=self)

    def on_stop(self):
        self.workspace.remove_core(core_id=self.core.id, is_assert=False, locust=self)


@events.init.add_listener
def on_locust_init(environment, **_kwargs):
    TimescaleListener(env=environment, testplan="9_add_table.py", target_env="myTestEnv")


if __name__ == "__main__":
    run_single_user(QuickstartUser, include_time=True, include_length=True, init_listener=on_locust_init)

[cyberw]

What do the Treelab methods call in the end? Is it http calls? If not, you need to make sure they are gevent-friendly, otherwise it will not work

[cyberw]

I had no issues running the following code. There might still be an issue in locust (as I think I've seen it sometimes) but maybe you have run into something that is more related to your particular setup..

...

class DemoUser(FastHttpUser):
    host = f"http://google.com/"
    wait_time = between(1, 1)

    def on_start(self):
        self.client.get("/?0")

    @task
    def t(self):
        with self.client.get("/?1") as resp:
            pass

    def on_stop(self):
        self.client.get("/?2")


@events.init.add_listener
def on_locust_init(*, environment, **_kwargs):
    listeners.TimescaleListener(testplan="demo", target_env=env, env=environment)

[masonCaminer]

I use the same principle as you, both functions in fasthttp.py

[masonCaminer]

You can see it. http://jenkins.treelab.work:3000/d/qjIIww4Zz/locust?viewPanel=15&orgId=1&from=1600071428567&to=1600071550332

[masonCaminer]

The username and password are all test

[masonCaminer]

locust -f src/loadTest/11_add_core.py --users=5 --spawn-rate=5 --headless --run-time 20s

[masonCaminer]

I only run 20 seconds, but after 20 seconds it keeps going

[masonCaminer]

USER_CREDENTIALS = PerformanceTest.get_users()
PerformanceTest.init_locust_env()
setup_logging(loglevel="INFO")


class QuickstartUser(FastHttpUser):
    wait_time = between(1, 1)
    host = os.getenv('host')

    def on_start(self):
        if len(USER_CREDENTIALS) > 0:
            name, password = USER_CREDENTIALS.pop()
            self.login = Treelab.login(email=name, password=password)
            try:
                self.workspace = Treelab.get_workspaces(token=self.login.token)[0]
            except:
                self.workspace = Treelab.add_workspace('压测', token=self.login.token)

    @task
    def add_core(self):
        self.core = self.workspace.add_core(str(random.randint(1, 100)), locust=self)

    def on_stop(self):
        self.workspace.remove_cores()


@events.init.add_listener
def on_locust_init(environment, **_kwargs):
    TimescaleListener(env=environment, testplan="11_add_core", target_env="myTestEnv")

[masonCaminer]

You can see grafana and you can see two problems.The first is on_ start time is also included in the total time. The second problem is the above

[masonCaminer]

I'm in Slack @you

[cyberw]

I have replied now. But you really must try to reproduce this without using your custom client, otherwise I cannot help.