Pick a moment
Choose a source and a point in time. Each of its tables goes back to a load from before that moment.
Retention and point-in-time revert
When a source sends wrong data, its tables can go back to how they were before. SchemaVortex rebuilds them from the copy of every load it keeps in your own storage, while your reports keep running. At the other end, history older than the retention period you set is removed automatically.
In short
A broken export, a faulty migration or ransomware in a source system can reach the lakehouse before anyone notices. A revert takes it out again.
Choose a source and a point in time. Each of its tables goes back to a load from before that moment.
Before anything changes, a preview shows for every table what it goes back to and what will be removed, and lists the tables that cannot go back.
Reports keep reading the current state until the rebuilt tables take over in one step.
The tables are rebuilt from the copies of the loads kept in your storage, with no new read from the source system.
How it works
SchemaVortex keeps a copy of each load from a source for the retention period you set. In the product, a load is an Extraction and the Integration that folds it into the Vault. A revert removes the loads from the chosen moment on and rebuilds each table from the ones before it.
erpKept since 12 Mar, 12:00Revert to 14 Mar, 02:00ordersback to 13 Mar, 23:10invoicesback to 14 Mar, 01:00price_listcannot go backRetention removes the loads older than the period you set. A revert removes the newest ones when a load went wrong. The data between the two lines never changes.
The source has to be fixed, or its loads paused, before a revert. Otherwise the next load brings the bad data back.
Before a revert
The loads a revert removes cannot be brought back. The steps around it make the choice explicit.
Only a Vault Lifecycle Manager of that source can run a revert.
The source's name has to be typed to confirm, and every revert is recorded with who ran it.
The History tables show how the data looked at an earlier moment, and nothing is removed.
A short live demo: we load bad data, revert the source to the hour before, and the report shows the right figures again.