The fantastic WebPageTest, free to use and public, has been available to set up your own private instances for many years; I wrote this up a while back, and scripted a Terraform version to make this as easy and automated as possible.
For AWS it was just a case of creating an EC2 instance (other installation options are available) with a predefined WPT server AMI (amazon machine image), add in a few configuration options and boom – your very own, autoscaling, globally distributed, website performance testing solution! New test agents would spin up automatically in other AWS regions, all based on WebPageTest Agent AMIs.
In 2020 WebPageTest was bought by Catchpoint and we finally saw improvements being made, pull requests being closed, and the WebPageTest UI getting a huge update; things were looking great for WebPageTest enthusiasts! If you havent heard of Catchpoint before, they are a company who are all about global network and web application monitoring, so a good match for WebPageTest.
Unfortunately, however, this resulted in the handy WebPageTest server EC2 AMIs no longer existing. If you want your own private installation you now need to build your own WebPageTest server from a base OS. It can be a bit tricky, though it gives you greater understanding of how it works under the hood, so hopefully you’ll feel more confident extending your installation in future.
In this article, I’ll show you how to create a WebPageTest private instance on AWS from scratch (no AMI), create your own private WebPageTest agents using the latest and greatest version of WebPageTest, and wire it all up.
Over the past few years I’ve been setting up, scripting, rebuilding, enhancing, tweaking, and hacking private WebPageTest installations. These have been configured to run hundreds of thousands of tests, globally, each day, pushing those stats into analytics platforms, data lakes, and/or databases, for monitoring, reporting, and alerting.
This latest version – a.k.a. "WebPageTest: Catchpoint 2021 Edition" (at least, that’s what I’m calling it) – needs a lot more work to set up (no predefined image you can use), but looks sooooo nice! The improvements to the UI were long overdue, and putting Core Web Vitals front and centre really speeds up investigations.
There’s a fair amount to get through, so I’ll break down what we’re going to do into the 3 main sections:
💻 Setting up your WebPageTest Server:
- Installation script
tl;dr – start a VM, run the WPT Server install script, configure a service user with IAM permissions to start and stop EC2 instances.
🤖 Custom Agents:
- Installation Script
- Copy AMI
tl;dr – start a VM, run the WPT Agent install script, create an AMI, copy AMI to other regions.
- Server location config
- Server settings
tl;dr – add the WPT agent AMI IDs to the WPT server config, party time! 🎉
The first two sections can be mostly summarised as "create an instance, run a setup script", so if you’re comfortable with that sort of thing on AWS, you could skim over, skip to the Agent AMI section, or even the final "Connect" section.
Don’t skip all the way to the end as there’s some cool stuff about easily overriding the installed WebPageTest config without changing the files from the github install script:
common" to get straight to it! It’s really handy, trust me.
Let’s get started!
Part 1: 💻 Setting up your WebPageTest Server
Gone are the days when you could choose an AMI and have a fully functional WebPageTest private installation moments later. At time of writing we need to go through a few more steps to get the same result, but worry not – I’m here to lead you through them!
The main steps are:
- Launch Ubuntu EC2 instance 🚀
- Log in and run the WPT installation script 📜
- Grant permissions to create new WPT agent instances 👍🏻
Ready?.. Here we go!
Launch Ubuntu EC2 instance 🚀
Head over to your AWS console and navigate to EC2. We can whizz through the next bit as it’s mostly defaults.
- Select Ubuntu 18.04/20.04 (or whatever is currently the latest supported version; I use 20.04 with no issues)
- Select size (t2.micro is fine for now; you can always scale up; my setups tend to use m4.large)
- I’d suggest you update the security group to only allow access from your own IP for SSH (HTTP needs to be open for now, else the agents can’t poll for work!)
- Confirm keypair so you can SSH in; create a new one, or use an existing one
I’ll cover the details about locking down HTTP also in a future post; it’s a cool problem to solve, but great once you’ve done it. 🔐
Log in and run the WPT installation script 📜
Once your server is running, ssh in using:
ssh -i "<path to your keypair>" ubuntu@<public IP or DNS of your WPT instance>
Then run the command as specified on the WebPageTest installation repo:
bash <(curl -s https://raw.githubusercontent.com/WPO-Foundation/wptserver-install/master/ubuntu.sh)
Give that a few minutes to process everything and it’ll eventually give you this message:
Setup is complete. System reboot is recommended. The locations need to be configured manually in /var/www/webpagetest/www/settings/locations.ini The settings can be tweaked in /var/www/webpagetest/www/settings/settings.ini The location key to use when configuring agents is: blahblahlblah An API key to use for automated testing is: blahblahlblah
Let’s reboot, then we can roll up our sleeves and get into the good stuff:
Next up is making sure the WPT server has permission to start WPT agents:
Grant permissions to create new WPT agents 👍🏻
Since our new server will be creating WPT agents (EC2 instances) on our behalf, we’re going to need to set up a user with permissions to start new EC2 instances.
Head over to the IAM area in your AWS console and create a user (programatic access only), granting
AmazonEC2FullAccess, noting down the user
secret for use in a moment.
Now we need to add this new user’s information to the WPT server configuration; here comes a really handy tip about working with the WebPageTest configuration settings:
All the WPT configuration settings can be found in
ini files in
/var/www/webpagetest/www/settings/ – it can be a pain to try to edit these files since you could end up resetting them with an eroneous
git pull; the best solution is to place your override files in a subdirectory called "common" –
/var/www/webpagetest/www/settings/common/ – these override the values from the same files in the parent directory.
Make the override directory:
Create a new file to add in our extra goodies:
sudo nano /var/www/webpagetest/www/settings/ common/settings.ini
(Yes I said
nano, alright? You can
viyourself if you must, but I’ve been a
nanokinda guy since I stopped using
emacsin the late 90s. Leave me alone. Grr.)
Add this to your new
settings/common/settings.ini file, putting in your own IAM details from the previous step:
; User to start/stop agents ; Access Key ID from IAM user setup ec2_key=AKIATS67N7IMQ5HZJREQ ; Secret Access Key from IAM user setup ec2_secret=kOyC+nxwVS7fXCgR4GcKpOB2Fj2kXSk8M+m5G65h ; enable ec2 ec2_locations=1
That last line means your server will look at the default
settings/ec2_locations.ini file to populate the agent list.
You should now find loads of lovely test locations in your WPT server! Yay!
However, if you actually try to use one of those locations to kick off a test, you may find the agent AMIs no longer exist (as of Nov 2021 that is):
$ tail -f /var/www/webpagetest/www/log/error.log.* 20:29:45 - EC2:Launching EC2 instance. Region: eu-west-1, AMI: ami-03009a026570febbb, error: The image id '[ami-03009a026570febbb]' does not exist 20:31:07 - EC2:Launching EC2 instance. Region: us-east-1, AMI: ami-0ff1ec702aef8a505, error: The image id '[ami-0ff1ec702aef8a505]' does not exist 20:31:19 - EC2:Launching EC2 instance. Region: ca-central-1, AMI: ami-057a97e61c7780543, error: The image id '[ami-057a97e61c7780543]' does not exist 20:31:29 - EC2:Launching EC2 instance. Region: ap-northeast-1, AMI: ami-07a931524bd2cde14, error: The image id '[ami-07a931524bd2cde14]' does not exist 20:31:37 - EC2:Launching EC2 instance. Region: ap-south-1, AMI: ami-0307d8587d2ce3b95, error: The image id '[ami-0307d8587d2ce3b95]' does not exist
In case agent AMIs provided by the installation script exist by the time you’re reading this (I could easily be out of date, considering how quickly WPT development is now moving), you can consider your private installation complete at this point! The rest of the article can be for reference to set up your own customised agents and server configuration overrides.
Oh dear. Only one thing for it – we need to create our own WPT agents too!
May as well delete that
common/settings.ini; we’re going to build our own list – change it back to:
; Access Key ID from IAM user setup ec2_key=AKIATS67N7IMQ5HZJREQ ; Secret Access Key from IAM user setup ec2_secret=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
Roll those sleeves up; we’re going to create our own WebPageTest Agent!
Part 2: 🤖 Custom Agents
Since the Amazon Machine Images (AMIs) in the default
settings/ec2_locations.ini don’t exist, we need to create our own WPT Agent AMIs to populate our own locations override file in
In order to do this, first we need to create one working WPT agent, and then use that as a template for the AMI.
Still with me? Sure you are! Let’s get cracking!
The main steps here are:
- Launch Ubuntu EC2 instance for the Agent 🚀
- Log in and run the WPT Agent installation script 📜
- Create AMI(s) 🤖🤖🤖
Launch Ubuntu EC2 instance for the Agent 🚀
- Head over to AWS EC2
- Select Ubuntu 18.04/20.04 (or whatever is currently the latest supported WPT agent OS version; I use 20.04 without issue)
- Select size (c4.large is recommended, but is not free so be aware)
- Confirm keypair so you can SSH in; create a new one, or use an existing one
Log in and run the WPT Agent installation script 📜
Once your instance running, ssh in and paste the installation command below, changing the browsers you want to install in the options – be aware that this will mean only the browsers you choose will be installed in the AMI you create; the more browsers you choose, the slower the installation and the slower the startup time of any subsequent agent that uses this AMI, since it will automatically install browser updates upon EC2 launch before executing tests, unless you explicitly disable browser updates via the WPT_UPDATE_BROWSERS=n option.
I guess this gives some interesting possibilities for optimisation of the initial startup time of an agent; instead of having one agent with many browsers, maybe have one agent per browser. Hmm. I might test that out…
Be sure to include
WPT_CLOUD=ec2 as this means the agent configuration will be passed in via user data automatically from your WPT server when it spins up an instance, instead of hard coding it in the agent.
This is an example agent installation config for just Chrome:
WPT_CHROME=y / WPT_FIREFOX=n / WPT_EDGE=n / WPT_BRAVE=n / WPT_EPIPHANY=n / WPT_OPERA=n / WPT_VIVALDI=n / WPT_CLOUD=ec2 / bash <(curl -s https://raw.githubusercontent.com/WPO-Foundation/wptagent-install/master/debian.sh)
Kick off your installation script, go have a walk, and eventually you’ll come back to a completed installation, suggesting you reboot. Instead of rebooting we’re going to shut it down so we can set the user data ourselves and test that it works before continuing.
Shut the instance down, then set the user data; only the
wpt_server needs to be correct at the moment, but we will use the other values in a moment to test it.
Select "Edit user data" for your shutdown Agent instance:
Set the user data:
wpt_server=<your WPT server's public IP> wpt_location=<some value, e.g. Test> wpt_key=<some value, e.g. 12341234123412341234>
Start the agent back up, then log in to your WPT Server and check the agent is polling for work; this will take a few minutes since the agent will update both itself (OS) and the installed browsers upon startup – I use the following command for this check:
> tail -f /var/log/nginx/access.log > 184.108.40.206 - - [15/Nov/2021:28:56:08 +8808] "GET /work/getwork.php?f=json&shards=1&reboot=1&servers=1&testinfo=1&location=Test&pc=1p-172-31-28-289-172.31.20.209&key=123412341234123412 348ec2-i-ec1b67169df4d868&ec2zone-eu-west-1a&version-211182.284522&dns-127.0.0.53&freedisk-3.466&pminutes-6 HTTP/1.1" 208 31 "--"wptagent" 220.127.116.11 - - [15/Nov/2021:20:56:00 +0000] "GET /work/getwork.php?f=json&shards-1&reboot-1&servers-1&testinfo-1&location-Test wptdriver&pc-ip-172-31-20-209-172.31.20.209&key-12341234 1234123412348ec2=i-Dec1b67169df4d868&ec2zone=eu-west-1a&version=211102.294522&dn=127.0.0.53&freedisk=3.466&upminutes=6 HTTP/1.1" 200 31 "-" 'wptagent" 18.104.22.168 - - [15/Nov/2021:28:56:05 +8808] "GET /work/getwork.php?f=json&shards=1&reboot=1&servers=1&testinfo=1&location=Test_wptdriver&pc=1p-172-31-20-289-172.31.20.209&key=12341234 123412341234&ec2-1-Dec1b67169df4d86&ec2zone-eu-west-1a&version-211182.284522&dns-127.0.0.53&freedisk-3.466&upminutes 6 HTTP/1.1" 200 31 "-" "wptagent" 22.214.171.124 - - (15/Nov/2021:20:56:05 +0808] "GET /work/getwork.php?f-json&shards=1&reboot-1&servers-1&testinfo=1&location-Test&pc-ip-172-31-20-289-172.31.20.289&key=123412341234123412 34&ec2=i-Dec1b67169df4d868&ec2zone=eu-west-1a&version=211182.284522&dn=127.0.0.53&freedisk=3.466&upminutes=6 HTTP/1.1" 200 31 ".- "wptagent" 126.96.36.199 - - [15/Nov/2021:20:56:18 +8808] "GET /work/getwork.php?f=json&shards=1&reboot=1&servers=1&testinfo=1&location=Test_wptdriver&pc=1p-172-31-20-289-172.31.20.209&key=12341234 123412341234&ec2-i-ec1b67169df4d868&ec2zone-eu-west-1a&version-211102.204522&dns-127.0.0.53&freedisk-3.4668 pminutes-6 HTTP/1.1" 200 31 "-" "wptagent" 188.8.131.52 - - [15/Nov/2021:20:56:10 +8808] "GET /work/getwork.php?f=json&shards=1&reboot-1&servers=1&testinfo=1&location=Test&pc-ip-172-31-20-209-172.31.20.209&key=123412341234123412 34&ec2=i-Dec1b67169df4d868&ec2zone=eu-west-1a&version=211182.284522&dn=127.0.0.53&freedisk=3.466&upninutes=6 HTTP/1.1" 200 31 ".- "wptagent"
Looks like it can connect, but it’s a good idea to test this agent works as expected before we create images from it, so let’s do that next.
Test your new WPT Agent
The new agent is polling for work, but your server doesn’t know about the agent yet ("Who are you? Stop calling me!"); we need to tell the server about this agent.
Back in your WPT server, create an override file
settings/common/locations.ini (remember that files in
settings/common override those in
settings/ with the same name) and paste in the following:
[locations] 1=Test_Group ;this value must match the "group" section header below default=Test_Group ;ditto ; agent group [Test_Group] 1=Test ;this value must match the "agent" section header below label="Test Group" ; this label value is what appears in the dropdown default=Test ;this value must match the "agent" section header below ; agent [Test] ;must match the value of wpt_location in the user data browser=Chrome label="Test" key=12341234123412341234 ;must match the value of wpt_key in the user data
[Test] and the value for
key must match whatever you used in the user data for
wpt_key earlier, i.e.:
wpt_server=184.108.40.206 wpt_location=Test wpt_key=12341234123412341234
Save that file and check that it’s connecting to the server ("Oh, it’s you! I know about you..") via
This should be pretty quick, since the agent is already up and polling for work.
Kick off a test from your WPT server and make sure it completes as expected.
All good? Alrighty, let’s continue!
Create AMI(s) 🤖🤖🤖
Now that we have a working agent, we need to create a machine image from it that will be used as the template when automatically spinning up agents on demand.
Select your WPT Agent and choose Actions -> Image and templates -> Create image
Be sure to note down the AMI id every time you create an AMI; you can easily find these again later if necessary, under EC2 -> Images -> My images (for each region).
Be sure to shut down any running WPT Agents before continuing to the next part, else it can get really confusing!
Part 3: 🔌 Connect
Now we can change the override file
settings/common/locations.ini to use our new AMIs;
[locations] 1=AMI_Test_Group default=AMI_Test_Group [AMI_Test_Group] 1=AMI_Test label="AMI Test Group" default=AMI_Test [AMI_Test] browser=Chrome label="AMI Test" key=12341234123412341234 ; watch out - extra stuff below! ami=ami-blahblah size=c4.large region=eu-west-1
ami to an agent section means your WPT server will attempt to launch an instance of that
ami of the specified
size in the chosen
region when a test is queued up.
Head back to your WPT server, kick off a test, and check that a new WPT Agent is created in the region you specified (check your AWS EC2 console); again, be patient!
Once that test succeeded with the AMI version, we can now add entries for all your other regional AMI locations; remember to add a new "group" at the top:
[locations] 1=AMI_Test_Group 2=Other_AMI_Test_Group default=AMI_Test_Group
and then the details for that group futher down, with the appropriate region and ami:
.... [Other_AMI_Test_Group] browser=Chrome label="Other AMI Test" key=12341234123412341234 ami=ami-blahblah ; different AMI ID for this region size=c4.large region=us-east-1 ; different region
Save that file and let’s finish up.
Kicking the tires
Let’s head back to your WPT server in a browser; see your locations in the dropdown?
Select one and kick off a test; a new EC2 instance should appear, with your AMI details and WPT server user data, in the chosen region!
After a few minutes you should see this agent in the WPT UI, pick up the test, and give you a result.
The agent will automatically be terminated to save money.
Congratulations! You have a fully functional private WPT server with your own WPT agent AMIs! Go you! 🥳
A brief reminder
Start a VM, run the WPT Server install script, configure a service user with IAM permissions to start and stop EC2 instances
Start a VM, run the WPT Agent install script, create an AMI, copy AMI to other regions
Add the WPT agent AMI IDs to the WPT server config, party time! 🎉
Evolving your setup
You could customise this further: for example, I have a lighthouse extension installed on some of my agent AMIs that attempts to calculate the environmental impact of the website, and I have AWS Inspector installed on some of my agent images, for security compliance purposes.
There’s loads more we can do from here – keep an eye out for more articles here on WebPageTest ⚡
- Are your config changes not being picked up? They might be cached internally, so you might need to restart nginx (
sudo service nginx restart) or even the entire server (
- Are you trying to use HTTPS to access your server, even though you haven’t set this up? Make sure you’re using HTTP for now else the WPT instance will look like it doesn’t work.
- Did you restrict IP access when setting up the security group? You need to open up HTTP to all traffic (
0.0.0.0), else WPT agents can’t connect.
- Pretty much anything else should be covered in my previous articles, where I list a bunch of troubleshooting usual suspects.