Getting help
Who to ask
Section titled “Who to ask”Most problems are permissions or configuration, and those belong to your tenant administrator rather than to the platform.
| Your problem | Ask |
|---|---|
| No queues, no tokens, no storage | Your tenant administrator |
| A permission you need | Your tenant administrator |
| A bucket you cannot mount | Your tenant administrator |
| A node that is offline | Your tenant administrator |
| Something broken across every tenant | The platform operator |
| The interface itself misbehaving | Report issue |
Report issue
Section titled “Report issue”Report issue at the bottom of the navigation sends a report to the address configured for your installation.
Include:
- The job id. It is in the detail panel and in every token transaction line. Without it a report is a description; with it, it is a record someone can look up.
- What you expected, and what happened instead.
- The queue, if it is job-related.
- Anything from the logs, if the job reached the point of having some.
Before reporting
Section titled “Before reporting”Two checks resolve most of it:
- Troubleshooting is ordered by how often each cause is the real one.
- Roles and permissions explains where a permission actually comes from - which is nearly always a user group, not your account.
Administrators: diagnosing for someone else
Section titled “Administrators: diagnosing for someone else”Impersonate the user. A superadmin can act as a specific account and reproduce the problem with their exact permissions. It is far faster than reasoning about what a group grants.
For a tenant admin, the equivalent is checking the user’s group directly on Users and groups - membership, the add-job permission, GPU limits, and which queues and buckets the group grants.