View and configure system parameters

What are system parameters?

System parameters in RisingWave refer to the parameters that advanced users can use to adjust how internal components work in RisingWave.

Currently, these system parameters are available in RisingWave.

Parameter	Description
adaptive_parallelism_strategy	The strategy for adaptive parallelism. Can be modified by ALTER SYSTEM command. When set to Full, the job will use all available parallelism; When set to Bounded(n), the maximum parallelism is limited to n; When set to Ratio(r), the actual available parallelism will be multiplied by r (rounded down). All parallelism is still subject to the limits of max parallelism and resource groups. See Configure adaptive parallelism strategy for more details.
barrier_interval_ms	The time interval of the periodic barriers. Since v2.5.0, this setting can also be configured per-database for finer control over latency and resource usage. See CREATE DATABASE for details.
checkpoint_frequency	Specify the number of barriers for which a checkpoint will be created. The value must be a positive integer. Since v2.5.0, this setting can also be configured per-database to better accommodate diverse workloads. See CREATE DATABASE for details.
sstable_size_mb	The target size of SSTable.
parallel_compact_size_mb	This parameter, together with max_sub_compaction, controls the concurrency of individual tasks. If the data size is smaller than parallel_compact_size_mb, only a single thread is used to execute the compaction task. If the data size of an individual task exceeds parallel_compact_size_mb, multiple concurrent threads are started to complete the task. At this time, if the data size is N, then the total number of these threads is determined by dividing N by parallel_compact_size_mb. Additionally, each sub-task’s size cannot exceed parallel_compact_size_mb, and the total number of sub-tasks cannot exceed max_sub_compaction.
block_size_kb	The size of each block in bytes in SSTable.
bloom_false_positive	False positive rate of bloom filter in SSTable.
state_store	The state store URL.
data_directory	The remote directory for storing data and metadata objects.
backup_storage_url	The URL of the remote storage for backups.
backup_storage_directory	The directory of the remote storage for backups.
telemetry_enabled	Whether to enable telemetry or not. For more information, see Telemetry.
max_concurrent_creating_streaming_jobs	The maximum number of streaming jobs that can be created concurrently. That is, the maximum of materialized views, indexes, tables, sinks, or sources that can be created concurrently.
pause_on_next_bootstrap	This parameter is used for debugging and maintenance purposes. Setting it to true will pause all data sources, such as connectors and DMLs, when the cluster restarts. This parameter will then be reset to its default value (false). To resume data ingestion, simply run risectl meta resume or restart the cluster again. Using RECOVER to trigger ad hoc recovery will not resume the cluster.
enable_tracing	Whether to enable distributed tracing. This parameter is used to toggle the opentelemetry tracing during runtime. Its default value is false.
time_travel_retention_ms	The data retention period for time travel. Defaults to 0, which disables time travel. To enable this feature, set it to a a non-zero value.
per_database_isolation	Enables per-database failure isolation, so that failures such as streaming job errors or worker crashes in one database do not affect other databases. Defaults to true. For more details of database isolation, see Workload isolation and interaction.

How to view system parameters?

You can use the SHOW PARAMETERS command to view the system parameters, along with their current values.

The Mutable column indicates whether the parameter can be altered using the ALTER SYSTEM SET command after the system is running. t means it can be altered using the ALTER SYSTEM SET command while f means it cannot be altered using the command.

SHOW PARAMETERS;

           Name                         |     Value           | Mutable
----------------------------------------+---------------------+---------
 barrier_interval_ms                    | 1000                | t
 checkpoint_frequency                   | 1                   | t
 sstable_size_mb                        | 256                 | f
 parallel_compact_size_mb               | 512                 | f
 block_size_kb                          | 64                  | f
 bloom_false_positive                   | 0.001               | f
 state_store                            | hummock+memory      | f
 data_directory                         | hummock_001         | f
 backup_storage_url                     | memory              | t
 backup_storage_directory               | hummock_001/backup  | t
 max_concurrent_creating_streaming_jobs | 1                   | t
 pause_on_next_bootstrap                | false               | t
 enable_tracing                         | false               | t
 time_travel_retention_ms               | 0                   | t
 ...

How to configure system parameters?

Mutable and immutable parameters are configured differently.

You can configure mutable parameters using the ALTER SYSTEM SET command in psql.

The full syntax of the ALTER SYSTEM SET statement is:

ALTER SYSTEM SET parameter_name { TO | = } { value | 'value' | DEFAULT };

Immutable parameters need to be initialized in the CLI of the Meta Node. state_store and data_directory need to be initialized before starting a cluster.

To configure parameter settings in the CLI of the Meta Node, navigate to the directory where RisingWave is installed and run the following command:

meta-node --<parameter_name> <value>

For example, to initialize the setting of data_directory:

meta-node --data_directory "hummock_001"

As RisingWave reads system parameters at different times, there is no guarantee that a parameter value change will take effect immediately. We recommend that you adjust system parameters before running a streaming query after your RisingWave cluster starts.