// Benutzung · Steuerung
Reading the audit log
The hub keeps one audit log of who did what: location deletes, agent offboards, registration tokens minted, checkpoint rollbacks, control actions, backup downloads, restores, NATS drains and the log's own retention prunes. The Audit log page shows it. How it is written, and what each action records, is in the audit log.
Who sees it
The page needs the audit.view permission, held globally: a role
assigned for one location does not grant it. The built-in admin role has it;
operator and viewer do not. Without it the page is not in the menu, and
opening /audit directly says it is not available to you. An administrator
can give audit.view to a role on the Access page (see
managing users & roles).
Reading a row
The rows are newest first. Each shows:
- Time, in your local time (hover for the exact UTC time),
- Actor, the user who acted (
systemfor the hub's own work, such as a retention prune), and their Client IP as the edge reported it, - Service, the one that did the action, the Action (for example
location.delete,control.k8s), its Target and its Location, - Phase, and the Result of a finished action.
Every action has two rows with the same correlation id: a proposed row written before it is done, and a terminal row after it: completed, failed or rejected (refused, nothing done). A request refused before anything was proposed has only a rejected row. Failed and rejected rows are highlighted.
Open a row (the arrow on its left) to see its parameters (as JSON), its whole result and its correlation id. Show this action's rows shows only the rows with that correlation id, so the proposed row and how it ended sit together. A proposed row without a terminal row is an action still running, or one whose outcome could not be recorded: check the system before you retry it.
Filtering
Every filter matches exactly (no partial text): actor, action, target, location, service, phase and correlation id. From and Until limit the time range, in your local time (From included, Until not). Choose Apply; Clear filters shows everything again.
The filters are in the page's address, so you can bookmark or send a filtered view, for example everything one user did, or every failed location delete. A time range from such a link is kept exactly as the link has it when you change another filter; only a From or Until you edit is read as your local time.
The page reads the log a page at a time (100 rows by default, at most 500; Page size). Load more reads the next, older page with the same filters.
When it cannot be read
- "The audit log is unavailable right now": inventory-svc, which keeps the log, is not answering, or its audit database is not migrated yet. The hub refuses audited actions in that state too. Try again once inventory-svc is up.
- "You are not permitted to read the audit log": your roles no longer
hold
audit.viewglobally. - "The filters were refused": the reason follows, for example a time the hub could not read.
How long rows are kept
Rows older than the audit retention are deleted by the retention reaper
(400 days by default; 0 keeps them forever). An administrator sets it on
the Retention page, under Audit log; the Audit log page links there.
Each prune is itself a row in the log.
The control panel of a location keeps its own short history of that
location's control actions; it reads the same log, and needs control.view
at that location rather than audit.view.