SET BACKGROUND_DDL
command to run DDL commands in the background.
Syntax
- By default,
BACKGROUND_DDL
is set asfalse
to disable it, meaning that DDL operations will execute in the foreground. The DDL commands will be blocking and only return until the backfill process is completed. - When
BACKGROUND_DDL
is set totrue
, any subsequent DDL operations will be executed in the background, allowing you to proceed with other tasks.
Supported DDL operations
For sinks created under
background_ddl
mode, RisingWave treats them as created
because their downstreams immediately start receiving the backfill data. Therefore, SHOW JOBS
does not track their backfill progress.Persistence
For materialized views being created in the background, their table definitions persist while they are being created, even if errors occur during checkpointing. This allows the materialized view jobs to be recovered from where they left off before the failure. Their table definitions and fragments will only be dropped if the job is canceled. For materialized views being created in the foreground, their table and fragments will be cleaned up if checkpointing fails, if the cluster is restarted, or if the stream job is canceled. The key difference is during the creating phase of a materialized view. After a materialized view is created (i.e. backfilling has completed), both foreground and background materialized views are functionally the same.Background management
Monitor progress
You can monitor the progress of background DDL operations using the SHOW JOBS command. The exception is forsink jobs
. These are immediately marked as completed, since their output will be immediately visible.
Cancel jobs
Currently background_ddl only supportssink
and materialized views
. You can just call DROP SINK
or DROP MATERIALIZED VIEW
on the corresponding relation to cancel it.
Set concurrent jobs
The maximum number of concurrent creating streaming jobs can be adjusted using theALTER SYSTEM SET max_concurrent_creating_streaming_jobs
command.
For example, you can set the maximum concurrent creating streaming jobs to 4: