In the path to have more stable metrics in our synthetic testing we have been trying out mahimahi, mitmproxy and WebPageReplay to record and replay Wikipedia. For mahimahi we have used patched version fixed by Gilles over Benedikt Wolters HTTP2 version of https://github.com/worenga/mahimahi-h2o. With mitmproxy and WebPageReplay we use the default version. The work has been done in T176361.
We have put mahimahi on ice because it is too much of hack to get HTTP/2 to work at the moment and WebPageReplay works out of the box with HTTP/2. mitmproxy worked fine but offered no clear benefit over WebPageReplay.
Replaying vs non replaying
Let us compare what the metrics looks like comparing WebPageTest vs WebPageReplay (Chrome).
The current version run that collects the data for https://grafana.wikimedia.org/d/000000059/webpagereplay-drilldown is a Docker container with this setup:
Running on AWS (instance type c5.xlarge) we get stable metrics. We have tried running the same code on WMCS, bare metal and Google Cloud and in all those cases the metrics stability over time was at least 2 to 4 times worse than AWS. This difference remains unexplained and probably lies somewhere in AWS's secret sauce (custom hypervisor, custom kernel).
On desktop we can use 30 frames per second for the video and we get a metric stability span of 33 ms for first visual change. Which is 1 frame of accuracy, since at 30fps one frame represents 33.33ms. Speed Index's stability span is a little wider but still ok (less than 50 points but it depends on the content).
For emulated mobile, we can use 30 frames per second but we seen that it would also work with 60 fps but somewhere we will hit the limit of the browser and OS. We run the both desktop and mobile with 100ms simulated latency during the replays.
We run tests from three servers at the moment:
- wpr-mobile.wmftest.org - Run tests on emulated mobile for WebPageReplay and user journeys
- wpr-desktop.wmftest.org - Run tests on desktop with WebPageReplay for Chrome
- wpr-enwiki.wmftest.org - Run tests on enwiki with WebPageReplay for Chrome and emulated mobile
Access the servers with the pem file:
# Running emulated mobile tests ssh -i "sitespeedio.pem" firstname.lastname@example.org # Running desktop tests ssh -i "sitespeedio.pem" email@example.com # Running enwiki tests ssh -i "sitespeedio.pem" firstname.lastname@example.org
Setup a new server
Here are the details of our current setup. We currently run desktop and emulated mobile Chrome tests on a C5.xlarge VM on AWS using Ubuntu 18.
Install it manually:
- Install Docker and grant your user right privileges to start Docker.
- Create a config directory where we place the secrets (AWS keys etc):
- Clone the repo (in your home dir) with the tests:
git clone https://github.com/wikimedia/performance-synthetic-monitoring-tests.git
- Take a copy of the
/config/secret.jsonfile that exists on of the current running servers and add it to
Also make sure the script start on server restart. When you start the script you choose which tests to run, by pointing out one or multiple test directories. That means that starting the tests looks differently on different machines.
@reboot rm /home/ubuntu/performance-synthetic-monitoring-tests/sitespeed.run;/home/ubuntu/performance-synthetic-monitoring-tests/loop.sh THE_TEST_DIR
That will remove the run file and restart everything if the server reboots.
The last step is to create a welcome message to you when you login to the server. Run
sudo nano /etc/profile.d/greeting.sh
echo "This server runs tests testing Desktop Wikipedia using WebPageReplay" echo "Start: nohup /home/ubuntu/performance-synthetic-monitoring-tests/loop.sh TEST_DIR &" echo "Stop: rm /home/ubuntu/performance-synthetic-monitoring-tests/sitespeed.run && tail -f /tmp/sitespeed.io.log"
Make sure to change the TEST_DIR and the message match what you run on your server.
Setup AWS monitoring
When you create a new instance, you also need to setup monitoring on that instance on AWS. Setup alarms for network outgoing traffic (NetworkOut) and set the alarm if it is <= 0 bytes for 3 out of 3 points for 1 hour. Assign the alert to the email group Performance-alerts.
Start and restart
You start by giving it the folder to test. If we test all desktop tests on the same machine we do that with:
Start the script:
nohup /home/ubuntu/performance-synthetic-monitoring-tests/loop.sh desktopReplay &
Restart: First remove
/home/ubuntu/performance-synthetic-monitoring-tests/sitespeed.run and then tail the log and wait for the script to exit. Then start as usual.
You can find the log file at
/tmp/sitespeed.io.log. There you can find all log entries from sitespeed.io.
Upgrade to a new version
Checkout the Performance/Runbook/WebPageReplay#Update to new version.
We also run alerts on the metrics we collect from WebPageReplay. Checkout Performance/WebPageReplay/Alerts.
Add a new URL to test
All configuration files exists in our synthetic monitoring tests repo. Clone the repo and go into the tests folder:
git clone ssh://USERNAME@gerrit.wikimedia.org:29418/performance/synthetic-monitoring-tests.git cd synthetic-monitoring-tests/tests
All test files lives in that directory. WebPageReplay tests exists in four directories:
- desktopReplay - the bulk of all the test URLs that gets tested for desktop. All these tests runs on one machine.
- desktopReplayInstant - the desktops tests for the English Wikipedia, we try to keep this list as short as possible and run them as often as possible to get fast feedback.
- emulatedMobileReplay - the bulk of all the test URLs that gets tested for emulated mobile. All these tests runs on one machine.
- emulatedMobileInstantReplay - the emulated mobile tests for the English Wikipedia, we try to keep this list as short as possible and run them as often as possible to get fast feedback.
The directory structure looks like this, where each wiki has their own file. Each file contains the URLs that is tested for that wiki.
. ├── desktop │ ├── alexaTop10.txt │ ├── coronaVirusSecondView.js │ ├── desktop.txt │ ├── elizabethSecondView.js │ ├── facebookSecondView.js │ ├── latestTechBlogPost.js │ ├── latestTechBlogPostSecondView.js │ ├── loginDesktop.js │ ├── mainpageSecondView.js │ ├── searchGoogleObama.js │ ├── searchHeaderObama.js │ ├── searchLegacyObama.js │ ├── searchPageObama.js │ ├── searchPortalObama.js │ └── shared │ └── searchScriptFactory.js ├── desktopReplay │ ├── arwiki.wpr │ ├── awiki.wpr │ ├── beta.wpr │ ├── dewiki.wpr │ ├── eswiki.wpr │ ├── frwiki.wpr │ ├── group0.wpr │ ├── group1.wpr │ ├── nlwiki.wpr │ ├── ruwiki.wpr │ ├── svwiki.wpr │ └── zhwiki.wpr ├── desktopReplayInstant │ └── enwiki.wpr ├── emulatedMobile │ ├── alexaMobileTop10.txt │ ├── elizabethSecondView.js │ ├── emulatedMobile.txt │ ├── facebookSecondView.js │ ├── latestTechBlogPost.js │ ├── latestTechBlogPostSecondView.js │ ├── loginEmulatedMobile.js │ ├── searchGoogleObama.js │ └── searchPageObama.js ├── emulatedMobileInstantReplay │ └── enwiki.wpr ├── emulatedMobileReplay │ ├── arwiki.wpr │ ├── beta.wpr │ ├── dewiki.wpr │ ├── eswiki.wpr │ ├── group0.wpr │ ├── group1.wpr │ ├── jawiki.wpr │ ├── nlwiki.wpr │ ├── ruwiki.wpr │ ├── rwiki.wpr │ ├── svwiki.wpr │ └── zhwiki.wpr ├── webpagetestDesktop │ ├── webpagetest.beta.wpt │ ├── webpagetest.enwiki.wpt │ ├── webpagetest.ruwiki.wpt │ └── webpagetest.wikidata.wpt └── webpagetestEmulatedMobile └── webpagetestEmulatedMobile.wpt
Let have a look at svwiki.wpr, it contains of three URLs:
https://sv.wikipedia.org/wiki/Facebook https://sv.wikipedia.org/wiki/Stockholm https://sv.wikipedia.org/wiki/Astrid_Lindgren
If you want to add a new URL to be tested on the Swedish Wikipedia, you open svwiki.wpr, add the URL on a new line and commit the result. When the commit passed through the Gerrit review, that URL will be picked up on the next iteration automatically by the test agent.
Debug missing metrics
If metrics stops arriving in Grafana the reason can be two different things: Either something is wrong with Graphite or something is broken on the WebPageReplay server.
Let us focus on the WebPageReplay server. Log into the server and check of if any tests is running. Do that by running
If everything is ok it should look something like:
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES 548ccf495779 sitespeedio/sitespeed.io:15.4.0 "/start.sh --graphit…" About a minute ago Up About a minute sitespeedio
We start (and stop) the container for every new test so a container should have been created for maximum a couple of minutes ago. If the created is a couple of hours (or days) ago, something is wrong. The container is stuck, probably something happened with the browser. You can fix the test by killing the container:
docker kill sitespeedio
The root cause (that the container got stuck) is still there but restarting the test usually works. After you killed the container, wait a minute and check again that a new container is running by using
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES f31eadffbcf0 sitespeedio/sitespeed.io:15.4.0 "/start.sh --graphit…" 4 seconds ago Up 3 seconds sitespeedio
Check the job the coming hour to make sure it doesn't get stuck again.