chrispenny/silverstripe-webpagetest

Submit performance test requests to WebPageTest, and consume the results

Installs: 0

Dependents: 0

Suggesters: 0

Security: 0

Stars: 0

Watchers: 2

Forks: 0

Open Issues: 2

Type:silverstripe-vendormodule

dev-master 2020-03-12 19:37 UTC

This package is auto-updated.

Last update: 2024-12-15 05:51:53 UTC


README

The goal of this module is to give you a way to queue and view WebPageTest performance tests. This is still early days though - we have some basic functionality, but this definitely isn't replacing a paid for testing service (at least not yet).

Out of box

  • Provides Request and Service classes that you can use to queue and consume WebPageTest performance test results.
  • Provides dev/tasks that can be used to queue and consume WebPageTest performance test results on a schedule.
  • Basic ModelAdmin to view your submissions and WebPageTest performance test results.

Still to come

  • Reports!
  • Graphs?
  • Consume more of the performance test result data

Installation

$ composer require chrispenny/silverstripe-webpagetest

Getting an API key for WebPageTest

If you would like to use this module with the public API, then you will need to request an API key. There are no requirements for getting one, it's just to make sure that folks are not abusing the API.

https://webpagetest.org/getkey.php

Required configuration

There are different ways that you can use this module, but, assuming you just want to go ahead and use the provided /dev/tasks, you'll need to add two configuration settings before you can go any further.

  • key (the API key that you requested above)
  • url (the URL that you would like WebPageTest to run the test against)

So, out of the box, we're only running performance tests against the same, single, URL.

In these early stages of the module, you'll need to implement your own /dev/task if you wish you support testing multiple different URLs. See Using the Services for details.

Default configuration

By default, in _config/submission.yml, we have set a couple of default configs:

base_url: http://www.webpagetest.org/runtest.php
format: json

The base_url is the standard request URL for WebPageTest. The format that this module supports is json.

Recommended configuration

  • location (if you don't provide a location, then it will be random, this makes your ongoing tests a bit random, as you'll likely get different results from different locations)

Available locations can be found here: https://webpagetest.org/getLocations.php?f=html

Other configuration

Out of the box, the module supports you sending any/all of the currently supported GET params in WebPageTest's RESTful API.

Available configuration

Using the dev/tasks

There are two dev/tasks available for you to use.

  • ChrisPenny\WebPageTest\Submission\Task
  • ChrisPenny\WebPageTest\TestResult\Task

Cron support is not adding as part of this module, but you could set up a cron to (say) run the Submission\Task once a day (resulting in one new WebPageTest being requested per day), and you might want to set up TestResult\Task to run every hour or so.

A full description of what each task does is below.

Submission\Task

When run, this task will send a single GET request to WebPageTest requesting that they queue a test matching what you have specified in your configuration.

The documentation for the RESTful API on WebPageTest will show you what the default values are if one is not supplied: https://sites.google.com/a/webpagetest.org/docs/advanced-features/webpagetest-restful-apis

It's important to note that tests through WebPageTest do not necessarily get processed straight away. If you are using the free service, then your request will go into a queue, and it will be processed when you're at the front of that queue. The tests then also (of course) take a while to complete. This is why we have a separate Task for retrieving the data for the tests.

Once we receive the "OK" (or otherwise) from WebPageTest, we create a Test Submission record, which you can see in the ModelAdmin under /admin/performance-tests/.

There are three possible statuses for these Submission records. These indicate the status of the test itself.

  • Pending (WebPageTest is still processing our request - they haven't tested our URL yet)
  • Success (The test has been completed by WebPageTest, and we should have the results)
  • Failure (Something went wrong ...)

If you want one new performance test run each day, then you would want to set up a cron triggering this task to run once each day.

TestResult\Task

When run, this task looks at what Submissions are current set with a status of Pending. It then requests an update from WebPageTest on what the status of that test is.

There is (generally) three possibility:

  • You are still in the queue, and your test has not begun yet
  • Your test is currently running
  • Your test has completed

In the case of the first two, we will create (or update) a Test Result record, and set its data to indicate to you where in the queue you are, or that the test is underway.

In the case where the test has completed, we will populate the Test Result record with the result data that we (currently) care about.

There is so... SO, much data in these test results. So for now, we just have the basics being stored. I welcome any and all to add support for us to store for of the data from the result data.

If you are running a cron once a day to trigger a test submission, you may want to queue the retrieval of the result, either for some time after the submission takes place, or, you could just trigger this task to run (say) every 20 minutes. The task does not do anything if there are no Pending submissions.

Viewing your submissions and results

The ModelAdmin can be found under: /admin/performance-tests/

This will list out all of the Submissions that you have made.

If you click on one of your Submissions, you'll see the current status, and once there is a result, a Result tab will show up, where you can see the Result and any of its "Runs" (you may have requested that WebPageTest performs multiple runs, and by default, it also performs a "first view" and "repeat view" test).