Still running Selenium?

Selenium suite flaky or Grid nodes crashing? We clear the backlog.

Most Selenium suites that reach us have been working for years and have slowly turned noisy: a driver that no longer matches the Chrome major, a stale element that only appears on a Tuesday, or a Grid node that silently drops a session every third run. We work through the driver, the waits and the orchestration until green is honest and red is actionable.

  • Since 2015Fixing technology
  • FreeEstimates
  • US teamReal engineers

selenium.log

still flaky

What it is

Selenium at a glance

Selenium is the long-standing suite of tools for driving real web browsers under program control. It began as Selenium Core in 2004, grew into Selenium RC and then Selenium WebDriver, and became the W3C WebDriver standard in 2018. Selenium 4 shipped in 2021 with native WebDriver BiDi work underway.

The core components are Selenium WebDriver (the client bindings in Java, Python, C#, Ruby and JavaScript), Selenium Grid (the distributed session router for running tests across many browsers and machines), and Selenium IDE (a browser-based record and replay tool). All of it is open source under the Apache 2.0 license.

Type
Browser automation and testing
First released
2004 (Selenium Core), 2007 (WebDriver), 2021 (Selenium 4)
License
Apache 2.0, open source
Maintained by
Software Freedom Conservancy and contributors
Typically used for
End-to-end web tests, cross-browser testing, regression suites, Grid-based CI runs, legacy automation

The honest version

Selenium pros and cons

No technology is all upside. This is where Selenium earns its place and where it tends to cause trouble.

Where Selenium is strong

  • Every browser, every languageChrome, Firefox, Edge, Safari and Internet Explorer modes are reachable through one API, with official clients in every mainstream server language. Few test frameworks match that reach.
  • The W3C WebDriver standardSelenium's protocol is a published standard. Browsers implement it directly, which means the vendor and the test suite are reading the same spec rather than a vendor API.
  • Grid scales horizontallySelenium Grid routes sessions to nodes running specific browsers and versions, which lets a team fan out a thousand tests across a cluster without a bespoke harness.
  • Mature and stableTwenty years of production use means most failure modes are documented, most environments have a known recipe, and the bindings do not shift their shape every quarter.
  • Works against real browsersTests run against the browser your users actually use, including quirks and extensions, rather than a headless renderer that only approximates behavior.
  • Huge ecosystem of integrationsVendors, cloud grids, reporting dashboards and CI plugins all speak Selenium. Replacing Selenium often means replacing those integrations too.

Where Selenium bites

  • Flakes are a fact of lifeA suite that passes 97% of the time still fires five false failures a day on a thousand tests. The underlying model (poll the DOM, hope the state is stable) invites races that Playwright and Cypress address at the protocol level.
  • Driver and browser versions drift apartChromeDriver 124 does not drive Chrome 125. On a team that updates Chrome without pinning the driver, the suite breaks on exactly the days CI should be reassuring.
  • Waits are the engineer's problemExplicit waits, implicit waits and Thread.sleep are easy to mix badly. The result is slow tests that still fail intermittently and nobody wants to touch.
  • Grid is not operationally freeA production Grid is a distributed system: nodes, queues, container images, a router and a monitoring story. Teams underestimate what keeping it green actually costs.
  • Slower than modern alternativesPlaywright and Cypress run the same scenario measurably faster, and their APIs remove whole classes of race condition. On greenfield projects, Selenium is rarely the first choice today.
  • Debugging requires disciplineSession logs, browser logs, screenshots, network capture and video evidence all need to be wired explicitly. Without them, a failure on CI is a mystery by lunchtime.

Sound familiar?

Common Selenium problems we fix

The Selenium problems that bring people to us, and what you have once each one is fixed.

session not created: ChromeDriver only supports Chrome version 124

Driver and browser versions mismatch

A CI image that updated Chrome overnight while ChromeDriver stayed pinned, or a Selenium Manager invocation that cannot reach the download endpoint and falls back to a stale binary.

Driver pinned alongside the browser

StaleElementReferenceException

Elements go stale mid-test

A reference captured before a React or Vue rerender, or an element that gets replaced by a routing transition. The classic retry wrapper hides it; a correct wait eliminates it.

Waits reference the right lifecycle

element click intercepted

Clicks hit the wrong element

A sticky header, a cookie banner or an overlay covering the target when the test ran in a particular viewport. Pure coordinate clicks succeed; interception errors fire on the real event path.

Overlays dismissed, targets reachable

DevToolsActivePort file doesn't exist

Chrome refuses to start on CI

A headless Chrome in a container without the flags it needs (--no-sandbox, --disable-dev-shm-usage), or a /dev/shm too small for the renderer.

Container configured for headless Chrome

Grid node ran out of available slots

The Grid saturates under load

More tests being dispatched than the Grid has sessions for, with the queue growing until sessions time out. A sizing exercise against real concurrency usually fixes it.

Grid sized to real demand

script timeout

A step takes longer than the script timeout

A single JavaScript execution via execute_async_script that blocks on a fetch, or a page that stalls under the renderer timeout. The right answer is almost always to narrow the wait.

Timeouts matched to real behavior

invalid session id

Sessions drop mid-run

A Grid node restarted during the test, a crashing browser, or a Docker container killed by an OOM. The suite needs to report the drop clearly rather than cascade into unrelated red.

Dropped sessions reported, not hidden

Unable to locate element

Selectors silently drift

A CSS selector that depended on a generated class name, a text selector that broke on a translation, or an ID that was renamed by a refactor. Test IDs are not glamorous and they prevent this.

Selectors pinned to stable attributes

What we do

Selenium help, from flake triage to a stable suite

Call us about a single flaky test, or hand us the whole Selenium suite and the Grid behind it.

Rescue
A red CI pipeline, a Grid that keeps running out of sessions, or a driver mismatch after a browser update. We get it green and keep it that way.
Flake reduction
Replacing bad waits, fixing stale element patterns, pinning selectors to test IDs, and separating real bugs from noise so the dashboard means something again.
Grid work
Right-sizing a Selenium Grid, moving to Docker-based nodes, or evaluating a managed grid against the operational cost of running your own.
Selenium to Playwright migration
Porting suites where the trade-off makes sense, keeping the parts of Selenium that still earn their keep, and avoiding a big-bang rewrite that leaves both halves half-done.
CI integration
Reports, screenshots, video capture and retries wired into GitHub Actions, GitLab or Jenkins so a failure is actionable without SSHing anywhere.
Managed Tech
Driver and browser version pinning, Grid uptime monitoring, flake dashboards and a human to ping when a Chrome update breaks a suite overnight.

How it works

From first message to fixed

  1. Look Free estimate

    Tell us what's going wrong. We look at it, then tell you plainly what's wrong and what it will cost.

  2. Fix

    We fix the cause, not the symptom. If something is down, getting it back up comes first.

  3. Harden

    Backups, monitoring, updates and access, so the same problem can't quietly come back.

  4. Hand back or stay on

    You get a plain-language write-up. Or we stay on for hosting, upkeep and new work.

Often found together

Running Selenium alongside something else?

We work on the whole stack, so the problem can be wherever it actually is.

Questions

Selenium questions

Something else? Call+1 737 727 7327

Should we move from Selenium to Playwright?

Sometimes, not always. Playwright is faster, less flaky and easier to debug for most greenfield projects. For a suite with thousands of tests, deep integrations and tight CI budgets, a wholesale rewrite rarely pays back inside a quarter. We assess the suite honestly and recommend either migration, hybrid, or stay with Selenium with the flakes fixed.

Why does our Selenium suite keep failing after Chrome updates?

ChromeDriver tracks Chrome majors and must match. Most CI images update Chrome automatically while the driver is pinned or shipped by Selenium Manager with its own cadence. Pinning both together, or wiring Selenium Manager to always fetch the matching driver, is what makes the updates stop breaking the suite.

What causes StaleElementReferenceException?

A reference to an element that the browser has since removed or replaced, usually because a framework rerendered the surrounding tree. The fix is a wait that captures the element after the mutation settles, not a retry loop around a stale handle.

How many Grid nodes do we need?

It depends on your concurrency, your session duration and your retry policy. A quick calculation using the longest test and the busiest CI run usually sizes the Grid to within a factor of two. We then measure, right-size, and document how to scale it next time traffic doubles.

Is Selenium still a safe long-term choice?

Yes for stable, large suites, and yes for environments where cross-browser coverage under the W3C standard matters. For new projects without those constraints, Playwright or Cypress usually win on developer experience. Both answers are honest; neither is universal.

How much does Selenium help cost?

Getting a quote costs nothing. Share the failing tests, the driver versions and the Grid shape, and a priced plan comes back before any changes are made.

Free estimate

Selenium giving you trouble?

Send us the failing test, the driver versions and the CI log. A real engineer replies with a priced plan, and nothing moves until you sign off.

  • No charge to find out what it will take
  • If it's urgent, say so

Replies come from a person. By sending, you authorize us to email you back. No marketing.