TestingBot Test Errors and How to Fix Them
Reference for the most common errors TestingBot returns when an automated or live test ends unexpectedly. Each section explains what triggered the error, the most common causes, and how to fix it. Errors appear on the test detail page in the dashboard and in the REST API response.
Find your error message
Search this table for the message your test framework printed (some start with ERROR). Parts in angle brackets, such as <session id>, differ per test.
| Message contains | Section |
|---|---|
Waited too long for job to start |
Concurrency limit reached: test queued |
You are currently at the limit of the maximum allowed parallel + queued tests |
Concurrency limit reached: test queued |
TestingBot was unable to fulfill your request: <device> is only available as a physical device |
Physical device only |
TestingBot was unable to fulfill your request: <device> is not available as a simulator/emulator |
Device not available |
TestingBot was unable to fulfill your request: <browser> is not available |
Capability not found or browser unavailable |
TestingBot was unable to fulfill your request, most likely the device is currently in use |
Device in use |
TestingBot was unable to fulfill your request for {…} please try again later |
No matching machine became free |
Unknown browserName <name>, we only serve …Unknown platform <name>, we only serve …
|
Capability not found or browser unavailable |
Please specify a deviceName |
Please specify a deviceName |
You need to enter your real key and secret |
Placeholder credentials |
Please include your client_key and client_secret parametersThe key (<key>) you used to authenticate is incorrect
|
Authentication errors (401) |
The test with sessionId <session id> is currently unknown in our systemSession <session id> ended <n> s ago: …
|
Session unknown or already ended |
The command you just sent (…) has no session ID |
Command has no session ID |
VM has ended <host> |
VM has ended |
TestingBot is temporarily busy, please retryTestingBot is experiencing high load and could not allocate a node in time
|
Grid temporarily busy |
Test cancelled from the TestingBot dashboard |
Cancelled from the dashboard |
Request aborted during test start |
New session cancelled before test start |
Parse error - please make sure you supply correct JSONInvalid JSON in Post body
|
Invalid JSON |
<session id> timed out: no new command was received |
Test idle timeout |
<session id> timed out: the test exceeded its maximum duration |
Test exceeded max duration |
<session id> stopped (credits depleted) |
Minutes depleted |
Messages from the browser driver or Appium, such as no such element, come with a W3C error code. See W3C WebDriver error codes.
Minutes depleted
What it means
The test failed because your account does not have enough test minutes left. New tests cannot be started on TestingBot until minutes are restored.
How to resolve
See our pricing plans and upgrade your subscription to add minutes immediately, or wait for your monthly plan to reset. Free-tier accounts receive a fresh allowance each month. Enterprise plans can request additional minute pools mid-cycle.
Test exceeded max duration
What it means
The test was terminated because it exceeded the maximum duration allowed. The default cap is 1800 seconds (30 minutes).
Common causes
- The test is genuinely long. Consider splitting it into multiple smaller, atomic tests.
- Tests on the cloud can be slower than local runs because of network round-trips and remote VM startup.
- The test is stuck in an endless loop, repeatedly sending commands.
How to resolve
- Break the test into smaller pieces.
- Use the maxDuration capability to raise the cap if you legitimately need a longer session.
- Check for endless loops and add explicit timeouts to waits.
Test idle timeout
What it means
TestingBot did not receive a new command from your Selenium or Appium script within the idle timeout. The session is terminated to prevent it from running forever and consuming all your minutes. The default idle timeout is 120 seconds for Selenium and Appium sessions, plus a 15-second grace period before the session is ended, and 90 seconds for Puppeteer and Playwright sessions. The test detail page shows This test did not see a new command for ... seconds. Test timed out.
Common causes
- A missing
driver.quit()at the end of the test. The runner finishes but the session keeps the browser alive. - The test script crashed, was interrupted, or lost internet connectivity.
- A legitimately slow step that takes longer than the idle timeout between commands.
How to resolve
- Ensure the test calls
driver.quit()in a teardown orfinallyblock so it always runs. - Run the test from a stable network. CI runners with unstable egress are a common cause.
- Raise the timeout with the idleTimeout capability if you need more time between commands.
Appium sessions
For Appium tests on real devices, simulators and emulators, the default idle timeout is 120 seconds. TestingBot allows a 15-second grace period before it ends the session, so the dashboard usually reports about 135 seconds.
- If you set
appium:newCommandTimeout, TestingBot uses that value as the session's idle timeout, up to a maximum of 3600 seconds. When bothnewCommandTimeoutandidleTimeoutare set,newCommandTimeoutwins. - The idle timer only resets when a command returns. A command that never comes back, for example because the on-device agent crashed (see device connection lost) or WebDriverAgent hangs (see WebDriverAgent not responding), ends the session with an idle timeout even though your script is still waiting.
- A large implicit wait makes every failed element lookup block for that long. Keep implicit waits short (a few seconds) and use explicit waits for slow screens, so no single command comes close to the idle timeout.
- Keep the HTTP read timeout of your Appium client below the idle timeout. Your client then gives up on a stuck command first, reports a clear error in your test output, and can still call
driver.quit(). - Always call
driver.quit()in afinallyblock or your framework's teardown hook. A session that is never quit keeps the device busy until the idle timeout ends it.
New session cancelled before test start
What it means
Your test runner requested a new session, but it closed the connection before TestingBot could allocate a slot and return a session id. The hub records this as Request aborted during test start.
Common causes
- Client connection timeout too short. iOS and Android simulators can take 60 to 120 seconds to boot.
- Running more tests than your plan's concurrency limit. Queued tests still consume the runner's connection time.
- Grid-side capacity issues during peak times. Check the TestingBot status page.
How to resolve
- Raise the connection timeout in your test runner. For WebdriverIO use
connectionRetryTimeout: 210000(3.5 minutes). For Selenium HttpClient set the equivalent socket timeout. - Confirm you are within the concurrent-session limit on your plan.
- Subscribe to the status page RSS feed for incident notifications.
Concurrency limit reached: test queued
What it means
Every plan has a maximum number of tests that can run in parallel. When all of your parallel slots are in use, a new session request does not fail right away: TestingBot keeps it in a queue and starts it as soon as one of your running tests finishes.
- Up to twice your plan's parallel limit can wait in the queue. Requests beyond that are refused with HTTP 429 and the message
You are currently at the limit of the maximum allowed parallel + queued tests for your TestingBot plan. - A queued request waits for up to 390 seconds (6.5 minutes). If it has not started by then, it fails with
Waited too long for job to start. - Queued requests are picked up oldest first. A request that needs a browser or device that is not free yet does not block the requests behind it.
- Time spent in the queue is not billed. Billing starts when the session starts on a browser or device.
- Your test runner has to keep the connection open while the request is queued. If the runner gives up first, the request is removed from the queue and you see New session cancelled before test start.
Espresso, XCUITest and Maestro runs
Espresso, XCUITest and Maestro runs are not limited to the 390-second queue. TestingBot keeps retrying the sessions of the run until a parallel slot is free. A run that still has not started 6 hours after it was created is cancelled.
How to resolve
- Set your test runner's parallelism to your plan's parallel limit, so tests do not sit in the queue.
- Raise the connection timeout of your test runner so it can wait for a queued request to start (see session cancelled).
- If you regularly need more parallel tests, see our pricing plans.
429 Too Many Requests
What it means
TestingBot returns HTTP 429 in two situations:
- Starting sessions: you requested more sessions than your plan can run and queue at the same time. See concurrency limit reached.
-
REST API: your API key (or IP address) sent more requests than the REST API rate limit allows. The response includes a
Retry-Afterheader andX-RateLimit-*headers showing your remaining budget.
How to resolve
- For sessions, lower your test runner's parallelism to your plan's limit.
- For the REST API, wait the number of seconds in
Retry-Afterbefore retrying, and avoid polling endpoints in tight loops. WatchX-RateLimit-Remainingto slow down before you hit the limit.
Device connection lost (socket hang up, ECONNRESET)
What it means
On real devices, simulators and emulators, Appium does not drive the app directly. It forwards each command to an automation agent running on the device: the UiAutomator2 server on Android, WebDriverAgent on iOS. When that connection breaks while a command is running, Appium returns errors such as:
Could not proxy command to the remote server. Original error: socket hang up-
ECONNRESETorECONNREFUSEDin the original error -
instrumentation process is not running (probably crashed)(Android)
Common causes
- The on-device agent crashed or was stopped, for example because the device ran low on memory.
- The app under test crashed and took the automation session with it.
- The connection between the device and its host dropped for a moment.
- The script sends commands in a tight loop (for example polling for an element hundreds of times per minute), which overloads the agent.
How to resolve
- Retry the test. A one-off connection drop usually does not repeat.
- Open the test on the dashboard and check the Appium log and the device log (Android or iOS log) around the failing command. A crash of your app or of the agent is usually visible there.
- Replace tight polling loops with explicit waits that check every few hundred milliseconds.
- If the error keeps happening on the same device, contact our support team with the session ID.
iOS: WebDriverAgent stopped responding
What it means
WebDriverAgent (WDA) is the XCTest-based agent that Appium uses to drive iOS apps. When WDA hangs, commands take very long or never return. Your test then fails with a timeout from your client, with a connection error if WDA crashed, or with a TestingBot idle timeout if no command returns in time.
Common causes
- A system alert (permissions, notifications, an OS dialog) covers the app and blocks the action.
- The app never becomes idle, for example because of continuous animations or a spinner, so WDA keeps waiting before it acts.
- A command has to walk a very large view hierarchy.
getPageSourceand broad XPath queries (such as//*) on busy screens are the usual suspects. - WDA or XCTest crashed on the device.
How to resolve
- Prefer accessibility id,
-ios predicate stringor-ios class chainlocators over XPath. - Do not call
getPageSourcein loops or before every step. Use it for debugging only. - Handle alerts explicitly, or use the
appium:autoAcceptAlertscapability. See permission popups. - For apps that are never idle, lower WDA's idle wait with the Appium settings API, for example
driver.updateSettings({ waitForIdleTimeout: 0 }). - A navigation command (
driver.get) that does not finish within 90 seconds returnsWaited too long for page to load (90s). Check that the page loads quickly on that device, and use the TestingBot Tunnel for internal or staging sites. - Check the Appium log on the test detail page for WDA errors or crashes around the slow command.
Manual session ended due to inactivity
What it means
A Live (interactive) browser or device session was terminated because there was no mouse or keyboard input for an extended period. Live sessions are billed by the minute, so idle sessions are ended automatically to prevent runaway charges.
How to resolve
Start a new Live session when you are ready to continue. If you need a long-running pinned session for QA workflows, consider switching to an automated test with a deliberate maxDuration setting.
Capability not found or browser unavailable
What it means
You will see a CapabilityNotPresentOnTheGridException (or a similar "browser not found" message) when the browser, version and platform combination you requested is not currently offered on the TestingBot grid. The TestingBot hub reports it with one of these messages:
-
TestingBot was unable to fulfill your request: <browser> <version> on <platform> is not available. Please check the browserName, version and platform capabilities against the browsers and platforms TestingBot offers.The request fails straight away instead of waiting in the queue, because nothing on the grid can run it. -
Unknown browserName <name>, we only serve …followed by the browser names TestingBot accepts. -
Unknown platform <name>, we only serve …followed by the platform names TestingBot accepts.
How to resolve
- Check the live supported browser list for available combinations.
- Use the version aliases
latest,latest-1,latest-2when you do not need to pin to an exact version. TestingBot resolves them to the current top release. - Double-check the spelling and case of
platformNamevalues. Selenium 4 W3C usesWINDOWS,macOS,LINUX, etc.
TestingBot was unable to fulfill your request
What it means
No browser, simulator, emulator or physical device on the grid can run the capabilities you sent, or the one that can stayed busy. The text after unable to fulfill your request tells you which case applies. A browser version or platform that is not offered is covered under capability not found.
…is only available as a physical device
Full message: TestingBot was unable to fulfill your request: '<device>' (<platform> <version>) is only available as a physical device, but your capabilities do not request one.
The device you asked for exists on TestingBot only as real hardware, and your capabilities ask for a simulator or emulator (the default). The request fails right away instead of waiting in the queue. Add realDevice to run it on the physical device:
"tb:options": { "realDevice": true }
…is not available as a simulator/emulator
Full message: TestingBot was unable to fulfill your request: '<device>' (<platform> <version>) is not available as a simulator/emulator, and no matching physical device is currently connected.
Nothing matches the deviceName and version you sent, neither virtual nor physical. The request fails right away instead of waiting in the queue.
- Check the spelling of the device name and the version against the device list.
- Leave out the version, or use a wildcard or a regex for the device name, so that more devices match. See device name rules.
- If you meant a physical device, also add
"tb:options": { "realDevice": true }.
…most likely the device is currently in use
Full message: TestingBot was unable to fulfill your request, most likely the device is currently in use. Please try again later, or consider specifying a regex …
You asked for a physical device that exists, but it stayed busy while your request waited in the queue (up to 390 seconds), for example because another test was running on it.
- Use a regex or wildcard
deviceName, such as"iPhone.*"or"Pixel.*", so any free device of that kind can take the test. See device name rules. - Do not pin one exact device in many parallel tests: they wait for each other.
- Retry later. If a device you need is never free, check the TestingBot status page or contact our support team.
…please try again later
Full message: TestingBot was unable to fulfill your request for {<your capabilities>} please try again later.
Your request waited in the queue for 390 seconds and no browser or virtual device that matches it became free. If your capabilities are correct, this is a capacity problem on our side, not something in your test.
- Check the capabilities printed in the message for a typo or an unusual combination.
- Retry the test.
- If it keeps happening, check the TestingBot status page and contact our support team with the capabilities from the message.
Please specify a deviceName
What it means
Your capabilities set platformName to iOS, Android or tvOS, but no deviceName. Without one, TestingBot cannot tell which simulator, emulator or device to start, so the request fails before it is queued.
How to resolve
- Add
appium:deviceName, for example"iPhone 16"or"Pixel 9". - A wildcard or regex works too, see device name rules.
- For mobile browser tests,
deviceNameis required as well. See mobile browser testing.
A related message, Invalid platformName and automationName combination. tvOS requires XCUITest, means you asked for tvOS with an automation name other than XCUITest.
Authentication errors (401)
What it means
Your API key and secret were missing, malformed, or did not match the credentials on file. The grid rejects the session with HTTP 401 before any browser is allocated. The message tells you which:
-
Please include your client_key and client_secret parameters in your request to the testingbot.com grid: the hub found no credentials in the request at all. -
The key (<key>) you used to authenticate is incorrect. Please check your TestingBot credentials: the key is unknown or the secret does not belong to it. -
You need to enter your real key and secret, which you can obtain from the TestingBot.com member area.: the key or secret still looks like the placeholder from an example: a key that contains the lowercase wordkeyor a secret that containssecret, such asapi_keyandapi_secret. Replace both with your own values.
How to resolve
- Verify your key and secret in your account settings.
- Pass credentials inside
tb:options(W3C or Selenium 4), not as top-level capabilities. - If you have rotated keys, update every CI secret store that references the old pair.
Session unknown or already ended
What it means
Your client sent a command for a session that TestingBot no longer runs. The hub answers with HTTP 404 and the W3C error invalid session id, and a message such as:
Session <session id> ended 42 s ago: no new command was received for 135 s (idleTimeout 120 s).
The test with sessionId <session id> is currently unknown in our system.
It may have already finished running.
When the hub knows why the session ended, the first line gives the reason. Without it, only the second part is shown: the session ended some time ago, or the id never belonged to a TestingBot session.
| Reason in the message | What to do |
|---|---|
it was ended by the client (quit) |
Your code called driver.quit() and then sent another command. Move the remaining steps, such as screenshots or log collection, before the quit. |
no new command was received … (idleTimeout …) |
See test idle timeout. |
the test exceeded its maximum duration … (maxDuration) |
See test exceeded max duration. |
the account ran out of test minutes |
See minutes depleted. |
it was stopped from the TestingBot member area |
Someone on your account stopped the test from the dashboard. |
the manual session's browser page showed no activity …the manual session was ended
|
See manual session ended due to inactivity. |
it exceeded the bandwidth limit |
The session transferred more data than allowed. Check the test for very large downloads or uploads. |
the node running it … (disappeared, stopped responding, was removed or was restarted)the hub could not reach the node running it
|
The machine running your test failed. This is not caused by your test: retry it, and check the status page if it happens repeatedly. |
How to resolve
- Check that you send commands to the session id your test created, and not to an id from an earlier run.
- Open the test on the dashboard: its error message shows why it ended.
Command has no session ID
What it means
Full message: The command you just sent (<method> <url>) has no session ID. The hub received a WebDriver command whose URL does not contain a session id, so it cannot tell which test it belongs to.
Common causes
- The test framework runs a step after the session has ended, for example a screenshot or a log download in a teardown method that runs after
driver.quit(). - A custom HTTP client or wrapper builds the command URL without the session id.
How to resolve
Move every step that talks to the browser or device before driver.quit(), and make quit the last step of the teardown.
VM has ended
What it means
Full message: VM has ended <host>. Your session still exists, but the virtual machine or device host that was running it is no longer registered with the grid, so the command cannot be delivered. This is not caused by your test.
How to resolve
- Retry the test.
- If it happens repeatedly, check the TestingBot status page and contact our support team with the session ID.
Grid temporarily busy
What it means
You see one of these messages when starting a session:
-
TestingBot is temporarily busy, please retry(HTTP 503) TestingBot is experiencing high load and could not allocate a node in time, please retry shortly
The grid could not handle your request at that moment. This is on our side, not caused by your capabilities or your plan's limits.
How to resolve
- Retry the session request after a few seconds. Many test runners can retry session creation automatically.
- If it keeps happening, check the TestingBot status page and contact our support team.
Test cancelled from the TestingBot dashboard
What it means
Your session request was waiting in the queue, and someone on your account cancelled it from the dashboard before it started. The request ends with Test cancelled from the TestingBot dashboard. No test ran and nothing was billed.
How to resolve
Nothing to fix. Start the test again if it was cancelled by mistake.
Invalid JSON
What it means
The body of the request could not be read as JSON. Depending on where it failed, the message is
Parse error - please make sure you supply correct JSON (HTTP 400),
Invalid JSON in Post body: … or Invalid request body.
How to resolve
- Use an up-to-date Selenium or Appium client library instead of building the request by hand.
- If you send capabilities from a script, validate the JSON first. Unescaped quotes in a test name or build name are a common cause.
W3C WebDriver error codes
Selenium 4, Appium 2 and later, and other W3C clients receive errors as a JSON body with an error code and a message, and your client turns the code into an exception. Most of these come from the browser driver or Appium on the machine running your test, and TestingBot passes them through unchanged. The ones TestingBot itself returns are invalid session id (a session that has ended) and session not created (the machine could not start the session).
| Error code | HTTP | What it usually means |
|---|---|---|
element click intercepted |
400 | Another element, such as a cookie banner, overlay or sticky header, covers the element you clicked. Close it or scroll the target into view first. |
element not interactable |
400 | The element exists but is hidden, disabled or has no size. Wait until it is visible and enabled. |
insecure certificate |
400 | The page's TLS certificate is not trusted, often on staging sites. See insecure certificates. |
invalid argument |
400 | A command got a value of the wrong type or shape, for example an unknown capability value or a malformed URL. |
invalid cookie domain |
400 | You tried to set a cookie for a domain other than the one of the current page. Open a page on that domain first. |
invalid element state |
400 | The element cannot take this action in its current state, such as clearing a read-only field. |
invalid selector |
400 | The CSS or XPath expression is malformed. |
invalid session id |
404 | The session has ended or never existed. See session unknown or already ended. |
javascript error |
500 | A script you ran with execute script threw an error in the page. |
move target out of bounds |
500 | An action moved the pointer outside the viewport. Scroll first or use a larger screen resolution. |
no such alert |
404 | You tried to accept or read an alert that is not open. |
no such cookie |
404 | No cookie with that name exists for the current page. |
no such element |
404 | The locator matched nothing. Use an explicit wait, and check that you are in the right frame or window. |
no such frame |
404 | The frame you switched to does not exist (any more). |
no such shadow root |
404 | The element has no shadow root. |
no such window |
404 | The window or tab was closed. Switch to an open window handle. |
detached shadow root |
404 | The shadow root you held a reference to is no longer in the page. Find it again. |
script timeout |
500 | An asynchronous script did not finish within the script timeout. |
session not created |
500 | The browser or Appium could not start a session with these capabilities, for example a driver and browser version that do not match or an app that fails to install. The message has the details. |
stale element reference |
404 | The element was removed from the page or re-rendered after you found it. Find it again before using it. |
timeout |
500 | A page load or implicit wait took longer than its timeout. |
unable to capture screen |
500 | A screenshot could not be taken, for example while a native dialog was open. |
unable to set cookie |
500 | The browser refused to set the cookie. |
unexpected alert open |
500 | An alert or prompt is open and blocks the command. Accept or dismiss it first. |
unknown command |
404 | The driver does not know this command, often a client that is newer or older than the browser driver. |
unknown error |
500 | Something failed that has no specific code. Read the message, and the browser or Appium log on the test detail page. |
unknown method |
405 | The command exists, but not with this HTTP method. |
unsupported operation |
500 | The browser or platform cannot do this, for example a feature one browser does not implement. |
Errors that the TestingBot hub raises itself, for credentials, plan limits or a busy grid, use the HTTP status codes 401, 429 and 503 and carry the message described in the sections above.
More help
Status incidents are posted at the TestingBot status page (RSS feed available). For anything not covered above, contact our support team with your session ID. The dashboard URL of the failed test is the fastest way for us to investigate.