Reading the Lineage Graph
The Lineage tab shows a graph centered on the container you opened. Every connection that container participates in is reachable from this view, and you can explore further in either direction to see more of the pipeline.
Direction and the Focal Node
The graph reads left-to-right:
- Upstream appears on the left. These are the sources that feed the focal container.
- Downstream appears on the right. These are the consumers of the focal container.
The focal node is the container whose Lineage tab you opened. It sits in the middle and is highlighted so you can tell it apart from the surrounding nodes.
Nodes
Each node represents a container and shows:
- A container-type icon on the left, telling a table, a file, and the three kinds of computed asset apart. It is tinted on the focal node, which also carries a coloured bar across the top of the card.
- The container name, or its display name when it has one. Hovering shows both the display name and the source name, so a renamed asset can still be matched to what the datastore calls it.
- A go-to arrow beside the name, which leaves the graph and opens that container's own page.
- The datastore name under the container name, so cross-datastore graphs are easy to navigate, with the datastore-type logo at the right of the header.
- An anomaly badge: an alert icon when the container has active anomalies, a check when it has none. A container that has never been scanned shows no badge at all, rather than a check it has not earned.
- A field summary reading
N of M fields linked, which is also the control that expands the field list. A container with no fields says so instead.
Assets Outside Qualytics
Lineage collected from a source datastore reaches tables that are not registered in Qualytics. Those appear as their own nodes, drawn with a dashed border and dimmed against the rest, carrying the logo of the datastore type the connection came from and the name the source spelled the asset with. Hovering names it in full, and right-clicking copies that name.
An external node is a reference, not an asset: there is no container behind it, so nothing opens from it and it has no fields to expand. Registering the table in Qualytics and running the collection again turns it into an ordinary node.
Edges
A badge on each edge shows how the connection was created: manual, data catalog, source collection, or Qualytics managed. See Lineage Sources for details.
Where a Connection Came From
Every connection can account for itself. Opening an edge's provenance shows what Qualytics read and what it derived the relationship from, so a connection you did not expect can be traced rather than guessed at.
A connection collected from a source datastore reports, for each piece of evidence behind it:
| Row | What it holds |
|---|---|
| Read from | The relation the evidence was read from, such as SNOWFLAKE.ACCOUNT_USAGE.ACCESS_HISTORY |
| Evidence | Recorded query history, Parsed query text, Parsed view definition, or Native lineage snapshot |
| Last seen | How long ago the relationship was last observed |
| Last sync | The Sync operation that last observed it |
Each piece is marked with the logo of the system that computed the relationship, which is the source itself when the evidence came from its own lineage, and the Qualytics symbol when Qualytics derived it by reading statement text. One connection can carry several, and each is listed separately rather than merged.
A connection with no such evidence still accounts for itself from its category alone: added by hand, imported from the data catalog, parsed from a computed table's SQL, taken from a computed join's declared inputs or a computed file's declared source, or written by a materialize or scan operation.
Expanding the Graph
The graph starts with the focal node and its immediate neighbors on each side. You can explore further from any node when you need more context.
Nodes that have more connections available show expand controls on the relevant side:
- Expand loads one more step in that direction.
- Expand all follows that direction to its end rather than one step, up to 20 steps out from the node.
See Also
-
Field-level Lineage
Expanding field lists, field metadata, and the focal field workflow.
-
Examples
Enrolling a warehouse, a medallion pipeline across datastores, a remediation table, and a hop only a person can record.
-
Best Practices
Which source to lean on, how to run collection, and how to keep a graph you can trust.
-
Permissions
The add-on gate, what each role can do, and why a deleted connection can come back.
-
How It Works
The connection model, the two levels of granularity, the rules a connection has to obey, and how to find the containers that have one.
-
Lineage Sources
The four categories a connection can come from, and how each is created, updated, and removed.