Troubleshooting
Find the message you are seeing, exactly as the status page or the sign-in page shows it, and the fix next to it. The messages stay in English here so you can search for them; … stands for the part that changes, such as a name or an error.
7 min read
On the status pageCopy link to this section
Start-up stops for missing configurationCopy link to this section
Needs configuration missing … — set them in the hosting settings and restart Set up Step {n} of {total} Add {keys} as a secret. The workspace restarts after a secret changes; this page picks it up. Open your hosting settings and add it there as a secret. Still missing — checked just now.
The status page and the sign-in page walk through what is missing one step at a time. Your administrator adds each item it names in the hosting settings; the workspace restarts after a change, and Re-check picks it up. At least one model provider key is needed. See Model providers.
A key was generated for this bootCopy link to this section
… was generated for this boot Set it as a secret so stored credentials survive a restart. Stored credentials will be lost on the next restart
No permanent encryption key was set, so the workspace made one for this run. Credentials stored now are encrypted with it, and after a restart, or in a restored database, they cannot be read. Backups are uploaded unsealed meanwhile, because their key is derived from it. Every signed-in page and the operations page say so until it is set. Your administrator sets a permanent encryption key in the hosting settings and restarts the workspace.
Still startingCopy link to this section
Starting up… Taking longer than usual: {m} seconds so far. Restarting after an unexpected stop.
The workspace's services start one after another, and the page lists each step as it finishes. Wait: it moves to the workspace by itself once it is ready. If one step stays last for several minutes, read its line.
A start-up step failedCopy link to this section
Start-up failed … exited …: … Why the start failed Failed step: {s}
The step marked ✕ names the service that stopped and the end of what it printed. Fix what that output points at, then restart the workspace. A failed database update usually means the data and the version disagree: restart once, and if it repeats, restore an earlier snapshot from the operations page.
A hung service is being restartedCopy link to this section
hung: … missed health answers — restarting it
A service stopped answering its health check while still running, so it was restarted. Nothing is needed if it comes back. If it keeps happening, the operations page shows the restart count for each service.
The restore at start-up failed and backups are pausedCopy link to this section
… — automatic backups are paused so the good snapshot is kept Paused: the restore failed, so the good snapshot is being kept
The newest snapshot could not be loaded, so the workspace started with an empty database and stopped automatic backups so it does not overwrite the good copy. Common causes are a changed encryption key, which the backup key is derived from, or backup storage that can no longer be read. Fix it and restart. See Backup and restore.
Backups are offCopy link to this section
Off — data lives only as long as this container Connect backup storage to keep it across restarts.
Your administrator connects backup storage and restarts the workspace. See Turn backups on.
Backups cannot be written, or backup storage is publicCopy link to this section
Backups cannot be written Backup storage is public Anyone can list and download the snapshots. Make the backup repository private in its settings. Could not check whether backup storage is private
At every start the workspace writes one small check file to backup storage and reads whether storage is private. If it cannot write, the backup token can read but not write, or storage is full or needs billing: give the token write access or free up storage, then press “Check storage again” on the operations page. If storage is public, make the backup repository private in its settings. See Turn backups on.
The sign-in settings could not be readCopy link to this section
could not read the sign-in settings: … — only the original sign-in password works
Until the database answers, a password changed on the operations page and two-factor sign-in do not apply: sign in with the password set when the workspace was deployed. Restart once the database step shows ✓.
A service stopped being restarted after repeated crashesCopy link to this section
crash loop: …
The service exited 5 times within 10 minutes, so it is no longer restarted automatically and is marked degraded. The operations page shows the reason and the last 20 lines it printed. Fix what those lines point at, then press Restart for that service on the operations page, which clears the stop and starts it again.
No snapshot before a database updateCopy link to this section
…; migrating anyway
A schema update was about to run and the snapshot taken before it failed, so the update ran without one. Take one with Back up now on the operations page, and check the backup settings named in the message.
Disk space is running lowCopy link to this section
low disk space on …: …
Free space for the database, the Bot workspace or the temp directory fell below 10% of the disk and below 5 GB, or below 256 MB on any disk. A large shared disk with plenty left does not alert. The Disk panel on the operations page shows each one. Remove files you no longer need from the Bot workspace, or move to hardware with more storage. The alert fires again only after space has recovered and runs low once more.
The sign-in password is missing or weakCopy link to this section
Sign-in password is not set The sign-in password is weak Set a long, uncommon passphrase as the sign-in password.
Your administrator sets the sign-in password in the hosting settings, a long passphrase that is not a common one, and restarts the workspace.
The model key did not answerCopy link to this section
Model key check failed ({p}) Model key: not configured No key for the chosen provider, so nothing was checked.
Once per start the workspace sends one short request with the key of the configured provider. Check the key and its provider with your administrator. See Model providers.
On the sign-in pageCopy link to this section
These messages come from the operations sign-in page, which keeps its own password. For the workspace's own sign-in screen, where everyone signs in with their own account, see Signing in and adding people.
The sign-in page shows these in the language you chose; the English text is below.
Sign-in is not configuredCopy link to this section
Sign-in is not configured yet. An administrator needs to set the workspace password in the hosting settings and restart it.
Your administrator sets the workspace password in the hosting settings and restarts the workspace.
Too many sign-in attemptsCopy link to this section
Too many attempts. Try again in a few minutes. Too many wrong passwords. You can try again in {s}s.
After 10 wrong passwords or codes from one address within 10 minutes, that address is refused until the oldest attempt is 10 minutes old. Each wrong password also makes that browser wait a little longer before the next try, and the sign-in page counts the wait down. Wait, then try again with the right password.
Wrong passwordCopy link to this section
That password is not right.
Check it with the administrator. Once a password has been set on the operations page, that one works, and the password set when the workspace was deployed still works as the break-glass password.
Two-factor code needed or wrongCopy link to this section
Enter your authentication code. That code is not right. Codes change every 30 seconds; a recovery code works once.
Enter the current 6-digit code from your authenticator app; if codes keep failing, check that the device's clock is set automatically. Without the device, use one of your recovery codes. Each works once.
The server is still startingCopy link to this section
The server is busy starting up. Try again in a moment.
The sign-in check could not run yet. Wait a moment and try again; the status page shows how far start-up has got.
The server cannot be reachedCopy link to this section
Could not reach the server.
The workspace may be asleep, restarting or updating. Reload in a minute; the status page answers as soon as the workspace is up.
Last updated