Locust: ValueError: StatsEntry.use_response_times_cache must be set to True

[max-rocket-internet]

I see this on every locust worker running 1.2:

$ locust --worker --locustfile=main.py
[2020-08-20 11:02:50,637] C02TD0F6GTDX/INFO/locust.main: Starting Locust 1.2
Traceback (most recent call last):
  File "src/gevent/greenlet.py", line 854, in gevent._gevent_cgreenlet.Greenlet.run
  File "/usr/local/lib/python3.8/site-packages/locust/stats.py", line 766, in stats_history
    'response_time_percentile_95': stats.total.get_current_response_time_percentile(0.95) or 0,
  File "/usr/local/lib/python3.8/site-packages/locust/stats.py", line 553, in get_current_response_time_percentile
    raise ValueError("StatsEntry.use_response_times_cache must be set to True if we should be able to calculate the _current_ response time percentile")
ValueError: StatsEntry.use_response_times_cache must be set to True if we should be able to calculate the _current_ response time percentile
2020-08-20T09:02:50Z <Greenlet at 0x10803f6a0: stats_history(<locust.runners.WorkerRunner object at 0x10806da60)> failed with ValueError

Tried with very basic locust file, same issue:

from locust import HttpUser, TaskSet, task, between

class LoadTestTask(TaskSet):
    def on_start(self):
        pass

    @task(3)
    def get_token(self):
        pass

class LoadTest(HttpUser):
    tasks = [LoadTestTask]
    wait_time = between(0.1, 1)

Any idea @cyberw or @vstepanov-lohika-tix?

[cyberw]

This line of code was added in #1516. @taojy123 , any ideas?

[taojy123]

Let me see.
I am trying to reproduce the problem first.

[taojy123]

Sorry! In my development, I did not consider that the use_response_times_cache parameter of worker mode is False.
Solution to fix it:
Add two lines in stats_history function of locust/stats.py:

def stats_history(runner):
    """Save current stats info to history for charts of report."""
    while True:
        stats = runner.stats
        if not stats.total.use_response_times_cache:  # *** add this
            break  # *** add this
        r = {
            'time': datetime.datetime.now().strftime("%H:%M:%S"),
            'current_rps': stats.total.current_rps or 0,
            'current_fail_per_sec': stats.total.current_fail_per_sec or 0,
            'response_time_percentile_95': stats.total.get_current_response_time_percentile(0.95) or 0,
            'response_time_percentile_50': stats.total.get_current_response_time_percentile(0.5) or 0,
            'user_count': runner.user_count or 0,
        }
        stats.history.append(r)
        gevent.sleep(HISTORY_STATS_INTERVAL_SEC)

Should I make a new PR? @cyberw

[max-rocket-internet]

Should I make a new PR?

Yes 🙂

[cyberw]

I made a PR myself :)

[cyberw]

Because this was in a release I made an emergency fix in my own PR & new release (1.2.1, on its way out now). @taojy123 , can you make a test case that excercises this scenario?

[taojy123]

ok
I made the test case in https://github.com/taojy123/locust/commit/2cd724c97b099a4c4a7158b0ae16d4cf04e57d58
The code like this:

    def test_stats_history(self):
        env1 = Environment(events=locust.events, catch_exceptions=False)
        runner1 = env1.create_master_runner("127.0.0.1", 5558)
        env2 = Environment(events=locust.events, catch_exceptions=False)
        runner2 = env2.create_worker_runner("127.0.0.1", 5558)
        greenlet1 = gevent.spawn(stats_history, runner1)
        greenlet2 = gevent.spawn(stats_history, runner2)
        gevent.sleep(1)
        hs1 = runner1.stats.history
        hs2 = runner2.stats.history
        gevent.kill(greenlet1)
        gevent.kill(greenlet2)
        self.assertEqual(1, len(hs1))
        self.assertEqual(0, len(hs2))

[cyberw]

Nice! Can you make a PR?

[taojy123]

ok, I just woke up,I'll make a PR later.

[taojy123]

@cyberw Hi~ I made the PR, see https://github.com/locustio/locust/pull/1538