"What's the best way to set up number of workers in Playwright for parallel testing?"
Hey folks! 👋
So, I've been tinkering with Playwright and trying to figure out how to *set up number of workers in playwright* for max speed without nuking my machine.
From what I’ve seen, just slapping `--workers=4` (or whatever) in the CLI *kinda* works, but is there a smarter way? Like, should I match it to CPU cores? Or is there a sweet spot?
Also, does it even matter if my tests are I/O-bound vs CPU-bound?
Kinda confused rn, ngl. Would love to hear how y'all *set up number of workers in playwright* for your setups.
Thx in advance! 🚀
*(PS: pls no "it depends" answers... gimme some real-world numbers!)*
Hey! So I usually set up number of workers in playwright based on my machine's cores minus one. Like, if I've got 8 cores, I’ll do 7 workers.
But honestly, it *does* depend on your tests. If they’re heavy on I/O (like waiting for APIs), you can go higher. For CPU-heavy stuff, keep it lower.
I’ve also used `playwright.config.js` to tweak workers per project. Helps a ton if you’ve got mixed test types.
Check out Playwright’s docs on sharding too—might be useful!
I just go with half my CPU cores for the set up number of workers in playwright. Works fine for me, but YMMV.
Also, if your tests are flaky, more workers = more pain. Found that out the hard way lol.
For real-world numbers, I run 4 workers on a 6-core machine. Any more and my fans sound like a jet engine.
Playwright’s default is usually sane, but if you wanna tweak, start low and ramp up. Monitor CPU/memory tho!
Pro tip: Use `--workers=auto` and let Playwright decide. It’s smarter than you think!
But if you’re stubborn like me, match it to your cores *minus one* for breathing room. Also, avoid setting up number of workers in playwright too high if your tests share resources (e.g., same DB).
If your tests are I/O-bound, you can *probably* go wild with workers. But CPU-bound? Nah.
I’ve had success with `--workers=2 * cores` for API tests, but for E2E, I stick to cores - 1.
Playwright’s GitHub has some wild debates on this—worth a skim!
Honestly, just benchmark it. Run your suite with 2, 4, 8 workers and compare.
I found that set up number of workers in playwright beyond 6 didn’t help me at all—just added noise.
Also, `playwright test --list` can help gauge test duration before optimizing.
Thx everyone! 🙌
Tried `--workers=auto` and it’s *kinda* magic? Also bumped it to 6 on my 8-core machine and saw a nice speedup.
But now I’m curious—how do you handle worker setup in CI? Same logic? Or do you cap it lower?
(Also, big shoutout for the `os.cpus()` tip—didn’t know that was a thing!)
I’ve seen folks use `os.cpus().length` in the config to auto-set workers. Fancy, huh?
But yeah, for most projects, cores - 1 is the sweet spot. Unless you’re testing on a potato. Then godspeed.
If you’re in CI, check your provider’s docs. Some limit workers weirdly.
Locally, I do cores + 1 for I/O stuff. Controversial, but it works for me.
Playwright’s `--trace` can help spot bottlenecks too!
