UI 'stop' button does not stop test

[colmsnowplow]

Description of issue

Using the web UI, the 'Stop' button temporarily slows down requests sent, which then rise back to previous levels.

Stop button then disappears.

Running on AWS EC2, 3 slaves.

Expected behavior

Stop button is expected to stop sending requests completely.

Actual behavior

RPS temporarily drops then resumes back to its previous level

Steps to reproduce (for bug reports)

• Configure locust on EC2 with 3 slaves, simple GET request test scenario (4 different requests)
• Config of 100 users, 1 swarm rate (via web UI)
• Attempt to stop test using 'stop' button.

[ernestom]

I see this behavior constantly as well with any number of slaves.

[georgevreilly-stripe]

This is #981, which was fixed in #982. However, this fix has never been released.

@cgoldberg, I reverted to 0.9.0 because this behavior is causing me severe problems. Please make a new release.

Also, although 0.11.0 is available at PyPI, it's never been tagged in this repo as a release.

[tsykora-verimatrix]

I've the same problem with 0.11.0.
We hit one of our endpoints hard (30+ slaves) and after hitting stop button in the UI the tests went on causing problems to one of our services. We had to physically take down those slaves to stop the test.

[georgevreilly-stripe]

I'm running 16 slaves† and the tests just won't stop until I kill the master process. There's a separate watchdog process that monitors the target's health and it is unable to stop the tests when running under Locust 0.11.

† This "slave" terminology is really unfortunate. "Worker" would be descriptive and inoffensive.

[max-rocket-internet]

We have the same problem. Running on k8s we have to delete all master and slave pods to reliably stop the test. It's quite annoying 😅

[rafaelbco]

Also, although 0.11.0 is available at PyPI, it's never been tagged in this repo as a release.

I'm quite confused about this too. Also, there are not Changelog entries for 0.11.0

[cyberw]

Hi all! Have you tried the latest release? (0.12.1)

[heyman]

I believe this should be fixed in the latest release (fixed in #982 as mentioned by @georgevreilly-stripe )

† This "slave" terminology is really unfortunate. "Worker" would be descriptive and inoffensive.

Check out #220. I'd be happy to merge a PR that changes that terminology.

[zhenhuaplan]

Hello all, when I have more than one worker, the problem still exists with the latest version 2.13.0, I am using Python 3.6

[cyberw]

2.13 is not compatible with python 3.6 - how did you even manage to install it?

[zhenhuaplan]

Sorry, I remember wrong, my locust version is 2.8.4, Python version is 3.6.8. However, this problem still exists when I tried to use the latest python and locust versions before, which is probably caused by the communication between the master node and the worker node or other problems. This problem does not exist when a single worker executes