Why browsing never wakes an engine
The rail is read from a search index that a scheduled crawl keeps up to date, not from a live connection to your data. That means:- Expanding a node costs nothing and never starts a stopped engine.
- You can browse the full catalog while every engine on your tenant is parked (scaled to zero).
- An “engines parked” pill on the rail confirms this: it tells you the tree is unaffected, not that something is broken.
Names, types and comments come from the index, not from a live read of the table. Opening an object’s page never reads a row of your data and never starts an engine.
When the rail is empty
Indexing is off by default, so a brand-new tenant sees an empty rail that says nothing has been indexed yet. How you fill it depends on your role:- If you can manage query engines, a Turn on indexing button appears. One click switches indexing on for the plane and starts the first crawl; the rail then shows it building and fills in when the crawl finishes. You don’t need to visit the settings page first.
- If you can’t manage query engines, the empty state asks you to have an administrator turn indexing on instead of offering a button that wouldn’t work for you.
What you see is what you may query
The tree only shows objects your role is allowed to see. If an object is hidden by policy, it looks exactly like an object that does not exist: there is no separate “access denied” branch to click through, and no count of what was filtered out. If you expected something and don’t see it, ask an admin for access rather than assuming it was never created. A typed filter narrows one level at a time (for example, filtering inside a schema for its tables). It does not search deeper levels. PressCmd K (or Ctrl K on Windows and Linux) to search across every catalog, schema, table and column at once. See Search.
The object page
Click any table or view to open it in a tab beside the tree. The page has four tabs:
A masked column is listed, not hidden. You’ll still see its name and type, with a note that reads return the mask rather than the stored value. This is a preview of what to expect: the query engine is what actually enforces access on every query, not this page.
The “Governed metrics” tab lists the semantic models bound to this table or view and offers “Define a model on this object”, which opens the model editor with the binding pre-filled. If your tenant does not have the semantic layer entitlement, the tab stays visible but explains that the feature is not enabled instead of loading anything. See Models and metrics.
Two ways to move into SQL
From an object page you can:- Preview rows: opens a new query tab beside the object with a
SELECTalready filled in and runs it immediately, on an engine that already serves this catalog. - Open in SQL editor: opens the same seeded statement without running it.
Deep links
Every object has a stable, shareable URL of the form/explore/<catalog>.<schema>.<table> (columns extend it with a fourth segment). Pasting one to a teammate opens straight to that object’s page, provided they too have access to it.
Related
Search your data
Find any table, column, metric or past query from one box.
Table lineage
See where a table’s data comes from and what depends on it.