Docker-Compose Locust UI response time not showing.

[lek18]

Prerequisites

• [X] I am using the latest version of Locust
• [X] I am reporting a bug, not asking a question

Description

I am running the docker-compose option for spinning up the locust UI, but the Response graph time is not showing/updating properly. The statistics chart give the correct information, its just not reflected in the UI (See images).

Here is my docker-compose.yml and command docker-compose up -d --scale worker=2

version: '3'

services:
  master:
    image: locustio/locust
    ports:
     - "8089:8089"
    volumes:
      - ./:/mnt/locust
      - ./surveys:/surveys
    command: -f /mnt/locust/locustfile-element.py --modern-ui --master
    environment:
      - TOKEN
    networks:
      - element

  worker:
    image: locustio/locust
    volumes:
      - ./:/mnt/locust
      - ./surveys:/surveys
    command: -f /mnt/locust/locustfile-element.py --worker --master-host master
    environment:
      - TOKEN
    networks:
      - element

networks:
  element:
    external: true

Additional information:

1. locust is calling a local endpoint from another docker image. docker run -p 9000:9000 --name element-word-cloud-luminate-dev-local --network element --rm --platform linux/amd64 -it element-world-cloud-luminate:latest bash -c "source /miniconda/bin/activate && conda activate seldonenv && seldon-core-microservice Model --service-type MODEL"

2. Both local endpoint docker image and locust docker-compose are using the same docker network.

3. The UI only appears to have this issue when the other element-word-cloud-luminate-dev-local is hitting 100% cpu

4. Similar UI behaviour if I use the python sdk of locust.

5. The UI behaviour goes away when I point to cloud version of my local docker element-word-cloud-luminate-dev-local

Here is the docker stats:

CONTAINER ID   NAME                                    CPU %     MEM USAGE / LIMIT     MEM %     NET I/O           BLOCK I/O         PIDS
1fc1687f5110   element-word-cloud-luminate-dev-local   100.18%   966.4MiB / 7.761GiB   12.16%    1.14MB / 19.4kB   14.9MB / 12.3kB   25
5dd9bee036e2   locust-worker-2                         0.02%     35.33MiB / 7.761GiB   0.44%     146kB / 1.45MB    0B / 0B           3
ab3e081ca846   locust-master-1                         0.04%     41.64MiB / 7.761GiB   0.52%     1.1MB / 7.28MB    2.04MB / 0B       3
9028c3d5bbee   locust-worker-1                         0.03%     34.21MiB / 7.761GiB   0.43%     127kB / 303kB     0B / 0B           3

Docker settings: 2 CPUs, 8Gb, ** same issue occurs if i bump the cpus to 4 CPUs and 16 GB

Command line

docker-compose up -d --scale worker=2

Locustfile contents

import json
import logging
import os

import urllib3

from locust import HttpUser, between, task

class PayLoadNotFound(Exception):
    pass

urllib3.disable_warnings()

# This TOKEN is an ENV variable, its needed for ELEMENT deployment.
TOKEN = os.getenv("TOKEN")
# Change Payload Here if you want to test something else.
file_name = "payload_3.json"

# For docker compose version
docker_file_name = f"/mnt/locust/surveys/{file_name}"

# This works if you are using python SDK instead of Docker.
local_file_name = f"surveys/{file_name}"

if os.path.exists(docker_file_name):
    with open(docker_file_name, "r") as file:
        request_Object = json.load(file)
elif os.path.exists(local_file_name):
    with open(local_file_name, "r") as file:
        request_Object = json.load(file)
else:
    raise PayLoadNotFound("File not found in docker nor in local.")

# this is your endpoint from Element.
endpoint = "world-cloud-main/api/v0.1"  # noqa
if not TOKEN:
    # This points to the docker container name + endpoint - ensure to have the same container name. # noqa
    # DO NOT EDIT endpoint NAME
    endpoint = "http://word-cloud-main-local:9000/api/v0.1"

class QuickstartUser(HttpUser):
    host = endpoint
    wait_time = between(1, 2)

    @task
    def test_world_cloud_service(self):
        headers = {
            "Content-Type": "application/json",
            "Authorization": TOKEN,
        }
        data = {"jsonData": request_Object}
        logging.info(
            f"This is the length of my inputStringList,"
            f" {len(request_Object.get('inputStringList'))}"
        )
        self.client.post("/predictions", json=data, headers=headers)

    def on_start(self):
        """on_start is called when a
        Locust start before any task is scheduled"""
        self.client.verify = False

Python version

python 3.11

Locust version

'2.19.1'

Operating system

MacOs Ventura 13.6.3

[cyberw]

Hi! Can you reproduce this issue with a minimal locustfile? Also, you are not on the latest release. Can you update?

[lek18]

Hi @cyberw , thank you for your response.

I updated the image to latest (https://hub.docker.com/r/locustio/locust/tags)

I see that the average response line in the UI is updating properly, the 99th percentile is still exhibiting the same behaviour, ie not updating and keeping flat like before.

[cyberw]

Can you remove as much as possible https://stackoverflow.com/help/minimal-reproducible-example

i.e, see if it happens without using docker for locust, without distributed, basic locustfile, etc. Right now there are too many ”special” things about your case.

[lek18]

Hello @cyberw , I used locust without docker and a basic locust file, and i do not see any issues in the ui.

command ran: locust -f locustfile-element.py

locust package version: 2.20.1

import json

import urllib3

from locust import HttpUser, between, task

urllib3.disable_warnings()

file_name = "payload_1.json"
# This works if you are using python SDK instead of Docker.
local_file_name = f"surveys/{file_name}"
with open(local_file_name, "r") as file:
    request_Object = json.load(file)
endpoint = "http://localhost:9000/api/v0.1"

class QuickstartUser(HttpUser):
    host = endpoint
    wait_time = between(1, 2)

    @task
    def test_world_cloud_service(self):
        headers = {
            "Content-Type": "application/json",
        }
        data = {"jsonData": request_Object}

        self.client.post("/predictions", json=data, headers=headers, verify=False)

[cyberw]

Do you still get the cpu load warning?

[lek18]

No @cyberw. I believe using the latest package made it worked. Thank you.