Stabilizing a Selenium build
summary
Concrete fixes for flaky Selenium builds: standardize boxes with Chef/Puppet, avoid RDP and use VNC, move the mouse out of the IE window and maximize browser windows to stop weird DOM failures, parallelize to cut a 3‑hour build to ~15 minutes (goal 10), run ~1M tests/day and surface 1-in-100 flakes with a flake detector, cache external assets and gems locally and route traffic through a black hole proxy (em-proxy). Tools: early notification plugins for RSpec/Cucumber/Jasmine, screenshots per node, and the open‑source selenium grid extras that kills browsers, reboots Windows, and plans video recording.
AI-written summary · about
chapters (24)
- 0:00Introduction
- 1:00Meet Dima
- 1:17Ive been impressed
- 1:44Contact information
- 1:56I love my job
- 2:11My wife is happy
- 2:26My partner in crime
- 3:39flakiness
- 4:23red master
- 6:08ajax
- 7:43selfcontained tests
- 8:58early notification plugin
- 10:13flake detector
- 12:21flaky environment
- 13:41Windows tips
- 16:32Speed up the test
- 17:55Blackhole proxy
- 19:14Cash external dependencies
- 21:00Selenium grid extras
- 21:17What is Selenium grid extras
- 22:41Future plans for Selenium grid
- 26:12Rebooting Selenium nodes
- 26:49Demos
- 30:23Questions
related talks
31mSelenium Hacks
37mSelenium In the Enterprise: What went right, what went wrong (so far)
38mFreeing up Selenium
39mTesting web apps with Selenium
37mBuilding Quality with Foundations of Mud
47mHow to Un-Flake Flaky Tests- a New Hire's Toolkit
42mReliable and Maintainable Webdriver Tests
23mAdaptive Testing