Skip to content

Troubleshooting

Work down this list; it is ordered by how often each one is the answer.

Symptom Cause Fix
Queue selector is empty Your group has no queues, or lacks add jobs Ask an administrator. See Users and groups
QUEUE JOB greyed out No image, or runtime zero on an interactive job Fill them in
QUEUE JOB greyed out, form complete No add-job permission on that queue The form names the missing permission
“This queue has no node that can run a job right now” Every node in the queue is offline or disabled Another queue, or wait. System shows which
Refused with a token message Balance will not cover the estimate Fewer resources, shorter runtime, or a top-up
A command is required Non-interactive queue with an empty command Add one, or use an interactive queue
Invalid environment variables A row with a name and no value, or the reverse Complete or remove it
Duplicate mount target Two mounts on the same path Change one

Almost always the image. Open the job’s logs:

In the logs Means
pull access denied Private image with no registry selected. See Docker images
manifest unknown The image or tag does not exist. Check the spelling
Your command’s own error The command exited non-zero
Killed shortly after starting Out of memory. Request more, or use less

This is the one that looks most like a bug and is not.

“No storage is attached to the … queue, so there is nothing to mount here” means the queue has no storage link. Try another queue; if the right one is still wrong, an administrator has to attach the storage to the queue. See Storage and Queues.

If it was not on a mounted bucket, it is gone permanently. The container filesystem is destroyed when the job ends, however it ended.

There is no recovery. Next time, mount a bucket and write results there - see Files and storage.

The runtime expired. A job’s runtime is a hard stop.

Extend it while it is still running - once it has expired there is nothing to extend, and the container is already gone.

The browser terminal ends with the page. Run long work under tmux, or with nohup, or as the job’s command so it does not depend on a shell at all. See Terminal and SSH.

Check
Was the port declared when the job was created? Ports cannot be added later. Clone the job
Is the port open in the Proxy view? Declaring is not opening
Is the job running? Port actions are disabled otherwise
Is the process listening on 0.0.0.0? Binding to 127.0.0.1 is unreachable from outside the container
Is there an IP filter on it? It may be excluding you

No private-port option at all means the relay is in no-DNS mode, which is installation-wide.

Invitations, password resets and notifications all go through SMTP. If none of them arrive, the tenant’s outgoing mail settings are the thing to check - and an administrator has to do it.

Report issue at the bottom of the navigation sends a report to your installation’s support address. Include the job id - it is in the detail panel and in the token transaction line, and it turns a vague report into a specific one.