WebDriver Chrome session fails with SessionNotCreatedException even though manual Selenium Chrome test works

[schiporovidiu-eng]

I’m trying to use changedetection.io with Chrome via WebDriver on an Ubuntu server.

What I did
Installed google-chrome-stable.

Installed and verified chromedriver.

Manually tested a Selenium Python script that launches Chrome headless with:

--headless=new

--no-sandbox

--disable-dev-shm-usage

--disable-gpu

That manual Selenium test works correctly and successfully opens https://example.com, so Chrome + ChromeDriver are working fine on this machine.

changedetection configuration
In the UI, I selected:

WebDriver Chrome/Javascript via "http://localhost:9515"

In settings:

fetch_backend = html_webdriver

Environment variables:

WEBDRIVER_URL=http://localhost:9515

CHROME_BIN=/usr/bin/google-chrome

GOOGLE_CHROME_BIN=/usr/bin/google-chrome

What happens
changedetection.io starts normally.

When I click Recheck on a watch, the log reaches RemoteWebDriver.

Then it fails with:

text
SessionNotCreatedException: session not created: Chrome instance exited. Examine ChromeDriver verbose log to determine the cause.
What I already verified
The WebDriver endpoint is reachable.

ChromeDriver is running.

Chrome works when launched manually through Selenium with the same machine and a working headless flag set.

The issue is not a missing browser installation.

My suspicion
It looks like changedetection.io is launching Chrome with a set of capabilities / flags that are incompatible with this environment, even though the same Chrome setup works when I launch it manually.

What I’m trying to achieve
I want changedetection.io to use the same Chrome launch options that work in the manual Selenium test, ideally via configuration only. If that is not possible, it seems like the WebDriver/Chrome startup logic may need to be adjusted.

Summary
Manual Selenium test: works

changedetection.io + WebDriver: reaches ChromeDriver, but Chrome exits immediately on session start

Likely issue: launch/capability handling inside changedetection.io, not Chrome itself

[salvador-castro]

Root cause diagnosis + actionable steps

Your manual Selenium script works because you control the ChromeOptions passed to the driver. changedetection.io constructs its own ChromeOptions internally and sends them to ChromeDriver when requesting a new session — if those capabilities are incompatible with your environment, Chrome exits immediately and ChromeDriver surfaces it as SessionNotCreatedException: Chrome instance exited.

Most likely causes in your specific setup:

1. Running as root without --no-sandbox — Chrome refuses to run as root unless --no-sandbox is explicitly passed. changedetection.io may not be injecting that flag when connecting to an external ChromeDriver endpoint. This is the Update README.md #1 cause of this exact error on bare Ubuntu servers.
2. /dev/shm too small — Chrome uses /dev/shm for shared memory. On many VPS setups it's only 64MB. If changedetection.io isn't passing --disable-dev-shm-usage, Chrome crashes silently on launch.
3. Missing --headless flag — Without a display server, Chrome exits immediately if not launched headless.

Step 1: confirm the exact flags being sent
Enable ChromeDriver verbose logging:
chromedriver --port=9515 --verbose --log-path=/tmp/chromedriver.log
Trigger a recheck, then inspect /tmp/chromedriver.log and look for the goog:chromeOptions > args block in the session creation request. Compare it against what your working manual script sends — the missing flags will be obvious.

Step 2: patch the fetcher to inject the required args
Locate the fetcher file:

find / -path "*/changedetectionio/fetchers/webdriver_fetch.py" 2>/dev/null
# older versions:
find / -path "*/changedetectionio/content_fetcher.py" 2>/dev/null

Look for the ChromeOptions block and ensure these are present unconditionally:

chrome_options.add_argument("--no-sandbox")
chrome_options.add_argument("--disable-dev-shm-usage")
chrome_options.add_argument("--headless=new")
chrome_options.add_argument("--disable-gpu")

These are the same flags your working manual script uses. Restart changedetection after editing.

Additional info that would help narrow this down:

• The goog:chromeOptions capabilities block from the ChromeDriver verbose log during a failed session
• Whether changedetection.io is running as root (whoami)
• Output of df -h /dev/shm
• changedetection.io version (pip show changedetection.io)