Contribute to the DSpace Development Fund

The newly established DSpace Development Fund supports the development of new features prioritized by DSpace Governance. For a list of planned features see the fund wiki page.

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 12 Current »

(Experiences from the University of Zurich, IT, ZORA Repository:)

To test the performance of our services (Eprints, DSpace, ...) we recommend the open source load testing tool Locust - https://locust.io

We want to test the multiple and parallel download of static and dynamic pages, the upload, various workflow steps and multiple user logins.

As I have not found any example configurations for DSpace, here is a short guide.

Installation

Instructions according to https://docs.locust.io/en/stable/installation.html

  • Lokal (Win, Linux, Mac) vs. RHEL8 Server; both installed and tested. I prefer a ServerSide installation, because bigger hardware and therefore more pressure during load tests.

  • python 3.11

  • gcc-c++

  • pip3.11

  • open Port 8089

locust install: pip3.11 install locust

01 Simple scenarios ("Hello World" mode)

Configuration

Simply pass through the URL entered at the start in the input form ("Number of Users", "Spawn rate", "Host") and test it - e.g. test  the homepage of your DS7 repo

vi locustfile.py

from locust import HttpUser, task

class HelloWorldUser(HttpUser):

@task def

   hello_world(self):

      self.client.get("/")

more on configs read https://docs.locust.io/en/stable/writing-a-locustfile.html#writing-a-locustfile

Start Locust

read Command Line Options

Attention: Standard http vs. https; Certs and Keys

[locust]$ locust --tls-cert /etc/pki/tls/certs/SERVERNAME/cert.pem --tls-key /etc/pki/tls/certs/SERVERNAME/privkey.pem

[2024-01-31 14:40:56,153] INFO/locust.main: Starting web interface at https://0.0.0.0:8089

[2024-01-31 14:40:56,166] INFO/locust.main: Starting Locust 2.20.1

UI

call Browser https://SERVERNAME:8089/ and start loadtests. First easy with 1 user:


Screenshot 2024-01-30 at 10-37-32 Locust for locustfile.py.png
(Attention: use trailing slash at the end of 'HOST'!)


See charts, failures, statistics and many more...

Screenshot 2024-01-30 at 10-39-05 Locust for locustfile.py.png

02 Real (warning) test scenarios (static & dynamic content)

  • Homepage

  • SubPage Communities

  • Search

  • Login Process incl. Token Autentication

and these scenarios according to the storybook in locustfile.py with

  • 1 user

  • 10 parallel users

  • nnn parallel users

Preliminary work / tests

Before the various authentication procedures for login, session handling (credentials, tokens, cookies, etc.) and workflow steps (dashboard, upload, metadata edit) are run through with try & error, it is advisable to test the individual work steps via the APIs beforehand using CURL.

For e.g. Before login ba POST method, receive TOKEN via GET. Read (insert Link)

Configuration (“complex” Mode)

vi locustfile.py

# Jens Witzel, Uni Zuerich 2024/01/30

import requests

from locust import HttpUser, task, between


class HelloWorldUser(HttpUser):

    @task

        def browse_homepage(self):

             self.client.get("/")


    @task

        def browse_cummunity(self):

             self.client.get("/communities/dc75f221-6ec0-4ff2-a72b-46d11444daa4")


         @task

         def search_items(self):

               needle = "test"

               self.client.get("/search?q={needle}") # Replace with the actual endpoint for searching items


       @task 

        def view_item_details(self): 

                item_id = "e0da84e0-0d71-41b6-9aee-17d7b29edca6" # Replace with the actual item ID

                self.client.get(f"/items/{item_id}") # Replace with the actual endpoint for viewing item details


class DSpaceUser(HttpUser):

           wait_time = between(1, 3) # Random wait time between 1 and 3 seconds

           @task def login(self): # get token

           getauthncookies = self.client.get("/server/api/authn", headers={})

          dscookies = dict(getauthncookies.cookies.get_dict()) 

           csrftoken = dscookies['DSPACE-XSRF-COOKIE']

           # post login with credentials and token

           response = self.client.post("https://SERVER/server/api/authn/login", {"user": USER, "password": PASSWORD}, headers={"X-XSRF-TOKEN":csrftoken}, cookies={"DSPACE-XSRF-COOKIE":csrftoken})

         if response.status_code == 200:

            print("Login successful")

        else:

           print(f"Login failed with status code: {response.status_code}")



Notes

  •     2 types of users (anonymous, registered user), login with token handling
  •     further, sequential storybook possible: login, workflow, upload, edit metadata, logout

Grafic Output (10 User)

Grafic Output (100 User)

Error rate increases as soon as system is overloaded


03 Really hard test scenarios: Recording workflow steps in FireFox to generate script for Locust

Creating a Locust configuration file from scratch is sometimes difficult, especially when complex workflows and work steps need to be analysed. 

One solution is to record typical work steps in the browser and save them in HAR format (
 ). This, in turn, can be created with the tool har2locust (
 ) into locustfile.py configuration files and an analysis as described above is ready to go.

The procedure:

    read the notes on

    [locust]$ pip3.11 install har2locust

Record FF session:

    Open DS7 page in browser

    Menu / Tools for developers | right mouse button + Examine

    Network analysis tab, click on bin (delete) - from then on the recording of the activities runs

    Execute DS7 actions (Login, MyDSpace, ... Logout)

    In column "File" right mouse button "Save as Har-File" save the recording

    sftp Har-File => SERVER

    [locust]$ har2locust x1.har > locustfile.py

    [locust]$ locust --tls-cert /etc/pki/tls/certs/SERVERNAME/cert.pem --tls-key /etc/pki/tls/certs/SERVERNAME/privkey.pem







  • No labels

All content on the LYRASIS Wiki is licensed under the CC BY (Attribution) license, unless otherwise noted.