BATCH UPDATE

BATCH UPDATE allows to update records in large tables while minimizing the risk of lock invalidation and transaction rollback by weakening guarantees. Specifically, data updates are performed as a series of transactions for each partition of the specified table separately, processing 10 000 rows per iteration. Each query processes up to 10 partitions concurrently.

This query, like the standard UPDATE, executes synchronously and returns a status. If an error occurs or the client disconnects, the data update stops, and the applied changes are not rolled back.

The semantics are inherited from the standard UPDATE with the following restrictions:

• Supported only for row-oriented tables.
• Supported only for queries with implicit transaction control (NoTx mode or EmptyTxControl in SDK).
• Only idempotent updates are supported: expressions following SET should not depend on the current values of the columns being modified.
• The use of subqueries and multiple statements in a single query is prohibited.
• The RETURNING clause is unavailable.

Example

BATCH UPDATE my_table
SET Value1 = "foo", Value2 = 0
WHERE Key1 > 1;