Time Machine overview
Time Machine shows how a table looked in the past and what changed between two points in time. Use it to check a nightly job that wrote wrong data, or to show an auditor the data as it was on a given day.
Time Machine only reads data. You do not need special SQL, and you do not need to restore or copy the table.
How snapshots work
Every time data is written to a table, the Lakehouse saves a new snapshot. A snapshot is a fixed, read-only version of the whole table at that moment. Older snapshots stay available, so you can go back to them.
Each snapshot records when it was written, what kind of write it was (for example append or overwrite), and how many records it added.
What you can do
| Task | What it answers | Page |
|---|---|---|
| View the snapshot timeline | When did the table change, and how? | View the snapshot timeline of a table |
| Preview past data | What did the rows look like at that time? | Preview data from a past snapshot |
| Compare structure | Did the columns, partitions, files, or record count change? | Compare the structure of two snapshots |
| Compare data | Which rows were added, removed, or changed? | Compare the data of two snapshots |
A typical investigation
- Open the timeline of the table and find the snapshot where the numbers changed.
- Preview the snapshot just before it to see the data as it was.
- Compare the structure of the two snapshots. This is fast and shows if the schema or record count changed.
- Compare the data of the two snapshots to get the exact rows that changed.
Comparing data reads a lot more than comparing structure. Time Machine estimates the size first and asks you to confirm before it runs.
Which tables work
Time Machine works with tables in the Lakehouse table format, which keeps snapshots. Views and tables in other formats have no snapshots, so Time Machine cannot open them.
A new table with no snapshot yet shows an empty timeline. A table needs at least 2 snapshots before you can compare.
Your data access applies
Time Machine reads data with your own account. The same rules as in Query Editor apply: the tables you can read, the columns you can see, row filters, and masked columns from Data Security. Time Machine never shows more than you can query yourself.
Default limits
Limits protect the shared Lakehouse from very large reads. An administrator can change them.
| Limit | Default |
|---|---|
| Rows shown in a preview | 1,000 |
| Data read by one data comparison, estimated before it runs | 10 GB |
| Rows read by one data comparison | 50,000,000 |
| Differences returned by one data comparison | 10,000 |
| Time to run a data comparison after you see its estimate | 5 minutes |
| Time limit for one request | 5 minutes |
| Data comparisons you can run at the same time | 1 |
Who can use Time Machine
| Permission | Lets you | Roles that have it |
|---|---|---|
time_machine:read | View the snapshot timeline and preview past data. | portal_admin, data_engineer, data_analyst |
time_machine:metadata_diff | Compare the structure of two snapshots. | portal_admin, data_engineer, data_analyst |
time_machine:data_diff | Compare the data of two snapshots. | portal_admin, data_engineer, data_analyst |
See Portal roles and permissions.