Skip to main content

system.settings

Introduced or updated: v1.2.745

Stores the system settings of the current session.

SELECT * FROM system.settings;
NameValueDefaultLevelDescriptionType
acquire_lock_timeout3030DEFAULTSets the maximum timeout in seconds for acquire a lock.UInt64
aggregate_spilling_memory_ratio6060LOCALSets the maximum memory ratio in bytes that an aggregator can use before spilling data to storage during query execution.UInt64
allow_query_exceeded_limit00DEFAULTAllow queries to overshoot the configured memory limit, deferring error notification until memory contention is encountered.UInt64
auto_compaction_imperfect_blocks_threshold2525GLOBALThreshold for triggering auto compaction after write. Set to 0 to disable auto compaction.UInt64
auto_compaction_segments_limit33DEFAULTThe maximum number of segments that can be reclustered automatically triggered after write.UInt64
binary_input_formatutf-8utf-8DEFAULTControls how string literals are interpreted when inserted into BINARY columns (HEX, BASE64, UTF-8, or UTF-8-LOSSY).String
binary_output_formathexhexDEFAULTControls how BINARY columns are rendered (HEX, BASE64, UTF-8, or UTF-8-LOSSY).String
bloom_runtime_filter_threshold30000003000000DEFAULTSets the maximum number of rows for bloom runtime filter generation.UInt64
collationutf8utf8DEFAULTSets the character collation. Available values include "utf8".String
compact_max_block_selection10001000DEFAULTLimits the maximum number of imperfect blocks that can be selected during a compact operation.UInt64
copy_dedup_full_path_by_default00DEFAULTThe default value if table option copy_dedup_full_path is not set when creating table.UInt64
cost_factor_aggregate_per_row55DEFAULTCost factor of grouping operation for a data row.UInt64
cost_factor_hash_table_per_row1010DEFAULTCost factor of building hash table for a data row.UInt64
cost_factor_network_per_row5050DEFAULTCost factor of transmit via network for a data row.UInt64
create_query_flight_client_with_current_rt11DEFAULTTurns on (1) or off (0) the use of the current runtime for query operations.UInt64
data_retention_num_snapshots_to_keep00DEFAULTSpecifies how many snapshots to retain during vacuum operations. Overrides data_retention_time_in_days. If set to 0, this setting will be ignored.UInt64
data_retention_time_in_days11DEFAULTSets the data retention time in days.UInt64
date_format_styleOracleOracleDEFAULTSets the date format style (used by datetime functions). Available values: "MySQL", "Oracle".String
ddl_column_type_nullable11DEFAULTSets new columns to be nullable (1) or not (0) by default in table operations.UInt64
default_order_by_nullnulls_lastnulls_lastDEFAULTSet numeric default_order_by_null mode. Values: "nulls_first", "nulls_last", "nulls_first_on_asc_last_on_desc".String
disable_join_reorder00DEFAULTDisable join reorder optimization.UInt64
disable_variant_check00DEFAULTDisable variant check to allow insert invalid JSON values.UInt64
dynamic_sample_time_budget_ms00DEFAULTTime budget for dynamic sample in milliseconds.UInt64
enable_aggregating_index_scan11DEFAULTEnables scanning aggregating index data while querying.UInt64
enable_analyze_histogram00DEFAULTEnables analyze histogram for query optimization during analyzing table.UInt64
enable_auto_analyze11DEFAULTEnables automatically analyze after write. 0 for disable, 1 for enable.UInt64
enable_auto_detect_datetime_format00DEFAULTEnable auto-detection for non-ISO datetime formats. Works across functions, COPY, and VARIANT cast.UInt64
enable_auto_fix_missing_bloom_index00DEFAULTEnables auto fix missing bloom index.UInt64
enable_auto_materialize_cte11DEFAULTEnables auto materialize CTE. 0 for disable, 1 for enable.UInt64
enable_auto_vacuum00DEFAULTWhether to automatically trigger VACUUM operations on tables.UInt64
enable_backpressure_spiller00DEFAULTUse new backpressure spiller.UInt64
enable_block_stream_write11GLOBALEnables block stream write.UInt64
enable_bloom_runtime_filter11DEFAULTEnables bloom runtime filter optimization for JOIN.UInt64
enable_cbo11DEFAULTEnables cost-based optimization.UInt64
enable_compact_after_multi_table_insert00DEFAULTEnables recluster and compact after multi-table insert.UInt64
enable_compact_after_write11DEFAULTEnables compact after write (copy/insert/replace-into/merge-into), needs more memory.UInt64
enable_cse_optimizer00DEFAULTEnables common subexpression elimination optimization.UInt64
enable_decimal_sum_widening00DEFAULTAutomatically widen SUM arguments from Decimal(19..38, scale) to Decimal(76, scale).UInt64
enable_dio11DEFAULTEnables Direct IO.UInt64
enable_distributed_compact00DEFAULTEnables distributed execution of table compaction.UInt64
enable_distributed_copy_into11DEFAULTEnables distributed execution for the 'COPY INTO'.UInt64
enable_distributed_merge_into11DEFAULTEnables distributed execution for 'MERGE INTO'.UInt64
enable_distributed_pruning11DEFAULTEnable distributed index pruning.UInt64
enable_distributed_recluster11GLOBALEnable distributed execution of table recluster.UInt64
enable_distributed_replace_into00DEFAULTEnables distributed execution of 'REPLACE INTO'.UInt64
enable_dphyp11DEFAULTEnables dphyp join order algorithm.UInt64
enable_dst_hour_fix00DEFAULTTime conversion handles invalid DST by adding an hour. Accuracy not guaranteed (disabled by default).UInt64
enable_expand_roles11DEFAULTEnable expand roles when execute show grants statement (enabled by default).UInt64
enable_experiment_aggregate11DEFAULTEnable experiment aggregate (enabled by default).UInt64
enable_experiment_hash_index11DEFAULTExperiment setting: enable hash index (enabled by default).UInt64
enable_experimental_connection_privilege_check00DEFAULTExperiment setting: enable connection object privilege check (disabled by default).UInt64
enable_experimental_new_join11DEFAULTEnables the experimental new join implementation.UInt64
enable_experimental_procedure11GLOBALEnables the experimental feature for 'PROCEDURE'.UInt64
enable_experimental_rbac_check11DEFAULTExperiment setting: enable stage and udf privilege check (enabled by default).UInt64
enable_experimental_row_access_policy00DEFAULTExperiment setting: enable row access policy (disabled by default).UInt64
enable_experimental_sequence_privilege_check00DEFAULTExperiment setting: enable sequence object privilege check (disabled by default).UInt64
enable_experimental_table_ref00DEFAULTExperiment setting: enable table ref (disabled by default).UInt64
enable_experimental_virtual_column00DEFAULTEnables experimental virtual column.UInt64
enable_fixed_rows_sort11DEFAULTEnable fixed rows sort serialize.UInt64
enable_geo_create_table11DEFAULTCreate and alter table with geometry/geography type.UInt64
enable_group_by_column_first00DEFAULTResolve GROUP BY names to input columns before SELECT aliases. Disabled by default for compatibility.UInt64
enable_hive_parquet_predict_pushdown11DEFAULTEnables hive parquet predict pushdown by setting this variable to 1. Default value: 1.UInt64
enable_join_runtime_filter11DEFAULTEnables runtime filter optimization for JOIN.UInt64
enable_last_snapshot_location_hint11DEFAULTEnables writing last_snapshot_location_hint object.UInt64
enable_loser_tree_merge_sort11DEFAULTEnables loser tree merge sort.UInt64
enable_materialized_cte11DEFAULTEnable materialized common table expression.UInt64
enable_merge_into_row_fetch11DEFAULTEnable merge into row fetch optimization.UInt64
enable_mutation_block_id_repartition11DEFAULTEnable local block_id repartition before row fetch in join-based mutations to reduce duplicate block reads.UInt64
enable_new_copy_for_text_formats11DEFAULTUse new implementation for loading CSV files.UInt64
enable_optimizer_trace00DEFAULTEnables optimizer trace.UInt64
enable_parallel_multi_merge_sort11DEFAULTEnables parallel multi merge sort.UInt64
enable_parallel_union_all00DEFAULTEnable parallel UNION ALL. Default is 0, 1 for enable.UInt64
enable_parquet_page_index11DEFAULTEnables parquet page index.UInt64
enable_parquet_prewhere00DEFAULTEnables parquet prewhere.UInt64
enable_parquet_rowgroup_pruning11DEFAULTEnables parquet rowgroup pruning.UInt64
enable_planner_cache11DEFAULTEnables caching logic plan from same query.UInt64
enable_proxy_bloom_pruning00DEFAULTEnable bloom index pruning during PROXY lightweight route estimation. Disabled by default to keep routing cheap.UInt64
enable_prune_cache11DEFAULTEnable to cache the pruning result.UInt64
enable_prune_pipeline11DEFAULTEnable pruning pipeline.UInt64
enable_query_result_cache00DEFAULTEnables caching query results to improve performance for identical queries.UInt64
enable_refresh_aggregating_index_after_write11DEFAULTRefresh aggregating index after new data written.UInt64
enable_replace_into_partitioning11DEFAULTEnables partitioning for replace-into statement (if table has cluster keys).UInt64
enable_result_set_spilling00DEFAULTEnable spilling result set data to storage when memory usage exceeds the threshold.UInt64
enable_selector_executor11DEFAULTEnables selector executor for filter expression.UInt64
enable_shuffle_sort00DEFAULTEnable shuffle sort.UInt64
enable_sort_spill_prefetch11DEFAULTEnable asynchronous restore prefetch for spilled sort blocks.UInt64
enable_sort_spill_stream_regroup11DEFAULTEnable regrouping sort spill streams by domain before merge.UInt64
enable_strict_datetime_parser11DEFAULTWhen enabled, datetime functions only accept ISO 8601 formats. When disabled, falls back to best-effort parsing.UInt64
enable_table_lock11DEFAULTEnables table lock if necessary (enabled by default).UInt64
enable_table_snapshot_stats00DEFAULTEnable analyze table statistics for snapshots.UInt64
enforce_broadcast_join00DEFAULTEnforce broadcast join.UInt64
enforce_local00DEFAULTEnforce local plan.UInt64
enforce_shuffle_join00DEFAULTEnforce shuffle join.UInt64
error_on_nondeterministic_update11DEFAULTWhether to return an error when updating a multi-joined row.UInt64
external_server_connect_timeout_secs1010DEFAULTConnection timeout to external server.UInt64
external_server_request_batch_rows6553665536DEFAULTRequest batch rows to external server.UInt64
external_server_request_max_threads256256DEFAULTRequest maximum number of threads to external server.UInt64
external_server_request_retry_times88DEFAULTRequest max retry times to external server.UInt64
external_server_request_timeout_secs180180DEFAULTRequest timeout to external server.UInt64
flight_client_keep_alive_interval_secs00DEFAULTSets the interval in seconds between two flight TCP keepalive probes. 0 disables keepalive.UInt64
flight_client_keep_alive_retries00DEFAULTSets the number of TCP keepalive retries for flight connections before declaring the peer unreachable. 0 disables keepalive.UInt64
flight_client_keep_alive_time_secs00DEFAULTSets the idle time in seconds before a flight TCP connection sends keepalive probes. 0 disables keepalive.UInt64
flight_client_timeout6060DEFAULTSets the maximum time in seconds that a flight client request can be processed.UInt64
flight_connection_max_retry_times00DEFAULTThe maximum retry count for cluster flight. Disable if 0.UInt64
flight_connection_retry_interval11DEFAULTThe retry interval of cluster flight in seconds.UInt64
force_aggregate_data_spill00DEFAULTFor testing only. Aggregate data will be forcibly spilled to external storage if enabled.UInt64
force_aggregate_shuffle_modeautoautoDEFAULTFor testing only. Shuffle mode for aggregation. Options: 'auto', 'row', 'bucket'.String
force_eager_aggregate00DEFAULTForce apply rule eager aggregate.UInt64
force_join_data_spill00DEFAULTFor testing only. Join data will be forcibly spilled to external storage if enabled.UInt64
force_materialized_cte_spill00DEFAULTFor testing only. Materialized CTE data will be forcibly spilled to external storage if enabled.UInt64
force_sort_data_spill00DEFAULTFor testing only. Sort data will be forcibly spilled to external storage if enabled.UInt64
force_window_data_spill00DEFAULTFor testing only. Window data will be forcibly spilled to external storage if enabled.UInt64
format_null_as_str00DEFAULTFormat NULL as str in query api response.UInt64
geometry_output_formatGeoJSONGeoJSONDEFAULTDisplay format for GEOMETRY values. Values: "WKT", "WKB", "EWKT", "EWKB", "GeoJSON".String
group_by_shuffle_modebefore_mergebefore_mergeDEFAULTGroup by shuffle mode. 'before_partial' is more balanced, but more data needs to exchange.String
grouping_sets_channel_size22DEFAULTSets the channel size for grouping sets to union transformation.UInt64
grouping_sets_to_union00DEFAULTEnables grouping sets to union.UInt64
hash_shuffle_bytes_threshold41943044194304DEFAULTSets the max bytes threshold for hash shuffle block partition stream.UInt64
hash_shuffle_rows_threshold81928192DEFAULTSets the max rows threshold for hash shuffle block partition stream.UInt64
hide_options_in_show_create_table11DEFAULTHides table-relevant information such as SNAPSHOT_LOCATION and STORAGE_FORMAT at the end of SHOW TABLE CREATE.UInt64
hilbert_clustering_min_bytes107374182400107374182400DEFAULTSets the minimum byte size of blocks for Hilbert Clustering.UInt64
hilbert_num_range_ids10001000DEFAULTSpecifies the domain of range IDs in Hilbert clustering. A larger value provides finer granularity but may incur a performance cost.UInt64
hilbert_sample_size_per_block10001000DEFAULTSpecifies the number of sample points per block used in Hilbert clustering.UInt64
hive_parquet_chunk_size1638416384DEFAULTThe max number of rows each read from parquet to databend processor.UInt64
http_handler_result_timeout_secs240240DEFAULTSet the timeout in seconds that a http query session expires without any polls.UInt64
http_json_result_modedisplaydisplayDEFAULTControls how HTTP query JSON data is encoded. Values: "display", "driver".String
idle_transaction_timeout_secs1440014400DEFAULTSet the timeout in seconds for active session without any query.UInt64
inlist_runtime_bloom_prune_threshold6464DEFAULTSets the maximum number of values in an IN list for runtime block bloom pruning.UInt64
inlist_runtime_filter_threshold10241024DEFAULTSets the maximum number of values in an IN list for runtime filter generation.UInt64
inlist_to_join_threshold10241024DEFAULTSet the threshold for converting IN list to JOIN.UInt64
input_read_buffer_size41943044194304DEFAULTSets the memory size in bytes allocated to the buffer used by the buffered reader to read data from storage.UInt64
join_runtime_filter_selectivity_threshold1010DEFAULTSelectivity threshold (percentage) for bloom join runtime filters. Default 10 means 10%.UInt64
join_spilling_buffer_threshold_per_proc_mb512512DEFAULTSet the spilling buffer threshold (MB) for each join processor.UInt64
join_spilling_memory_ratio6060LOCALSets the maximum memory ratio in bytes that hash join can use before spilling data to storage. 0 is unlimited.UInt64
join_spilling_partition_bits44DEFAULTSet the number of partitions for join spilling. Default value is 4, meaning 2^4 partitions.UInt64
lazy_read_across_join_threshold1010DEFAULTSets the maximum LIMIT in a query to enable lazy read across joins. Setting it to 0 disables the optimization.UInt64
lazy_read_threshold10001000DEFAULTSets the maximum LIMIT in a query to enable lazy read optimization. Setting it to 0 disables the optimization.UInt64
load_file_metadata_expire_hours2424DEFAULTSets the hours that the metadata of files loaded with COPY INTO will expire in.UInt64
materialized_cte_spilling_memory_ratio6060DEFAULTSets the maximum memory ratio in bytes that materialized CTE execution can use before spilling data to storage. 0 is unlimited.UInt64
max_aggregate_restore_worker1616DEFAULTSets the maximum number of workers for aggregate restore.UInt64
max_aggregate_spill_level33DEFAULTMaximum recursion depth for the aggregate spill. Each level repartitions data into 4 smaller parts.UInt64
max_block_bytes5242880052428800DEFAULTSets the maximum byte size of a single data block that can be read.UInt64
max_block_size6553665536DEFAULTSets the maximum rows size of a single data block that can be read.UInt64
max_cte_recursive_depth10001000DEFAULTMax recursive depth for recursive CTE.UInt64
max_execute_time_in_seconds00DEFAULTSets the maximum query execution time in seconds. Setting it to 0 means no limit.UInt64
max_hash_join_spill_level11DEFAULTMaximum recursion depth for the hash join spill. Each level repartitions data into 16 smaller parts.UInt64
max_inlist_to_or33DEFAULTSets the maximum number of values that can be included in an IN expression to be converted to an OR operator.UInt64
max_memory_usage5153960755251539607552DEFAULTSets the maximum memory usage in bytes for processing a single query.UInt64
max_public_keys_per_user1010DEFAULTMaximum number of public keys allowed per user for key-pair authentication.UInt64
max_push_down_limit1000010000DEFAULTSets the maximum number of rows limit that can be pushed down to the leaf operator.UInt64
max_query_memory_usage2576980377625769803776LOCALThe maximum memory usage for query. If set to 0, memory usage is unlimited.UInt64
max_result_rows00DEFAULTSets the maximum number of rows that can be returned in a query result. Setting it to 0 means no limit.UInt64
max_set_operator_count1844674407370955161518446744073709551615DEFAULTThe maximum count of set operators in a query.UInt64
max_spill_io_requests88DEFAULTSets the maximum number of concurrent spill I/O requests.UInt64
max_storage_io_requests6464DEFAULTSets the maximum number of concurrent storage I/O requests.UInt64
max_threads88DEFAULTSets the maximum number of threads to execute a request.UInt64
max_vacuum_temp_files_after_query1844674407370955161518446744073709551615DEFAULTThe maximum temp files will be removed after query. Disable if 0.UInt64
max_vacuum_threads11DEFAULTSets the maximum number of threads to execute vacuum operation.UInt64
min_max_runtime_filter_threshold1844674407370955161518446744073709551615DEFAULTSets the maximum number of rows for min-max runtime filter generation.UInt64
nested_loop_join_threshold1000010000DEFAULTSet the threshold for use nested loop join. Setting it to 0 disables nested loop join.UInt64
network_policyDEFAULTNetwork policy for all users in the tenant.String
numeric_cast_optionroundingroundingDEFAULTSet numeric cast mode as "rounding" or "truncating".String
optimizer_skip_listDEFAULTComma-separated list of optimizer names to skip during query optimization.String
parquet_fast_read_bytes10485761048576LOCALParquet file with smaller size will be read as a whole file instead of column by column. Default value: 1MB.UInt64
parquet_max_block_size81928192DEFAULTMax block size for parquet reader.UInt64
parquet_rowgroup_hint_bytes134217728134217728DEFAULTHint bytes for dividing large parquet files into multiple rowgroups for reading. Default value: 128MB.UInt64
parse_datetime_ignore_remainder11GLOBALIgnore trailing chars when parse string to datetime.UInt64
persist_materialized_cte11DEFAULTDecides if materialized CTEs should be persisted to disk.UInt64
prefer_broadcast_join11DEFAULTEnables broadcast join.UInt64
prewhere_selectivity_threshold100100DEFAULTMaximum selectivity percentage for pushing row selection into remain-column reads during prewhere.UInt64
proxy_routing_modelstatisticsstatisticsDEFAULTControls how PROXY chooses a target table. Values: 'statistics', 'prefix'.String
purge_duplicated_files_in_copy00DEFAULTPurge duplicated files detected during execution of copy into table.UInt64
queries_queue_retry_timeout300300DEFAULTThe retry interval for query queue timeout. 0 if never retry.UInt64
query_flight_compressionLZ4LZ4DEFAULTFlight compression method. Values: "None", "LZ4", "ZSTD".String
query_out_of_memory_behaviorspillingspillingLOCALIf the query memory limit is exceeded, the system will enforce predefined actions. Values: "throw", "spilling".String
query_result_cache_allow_inconsistent00DEFAULTDetermines whether Databend will return cached query results that are inconsistent with the underlying data.UInt64
query_result_cache_max_bytes10485761048576DEFAULTSets the maximum byte size of cache for a single query result.UInt64
query_result_cache_min_execute_secs11DEFAULTFor a query to be cached, it must take at least this many seconds to fetch the first block.UInt64
query_result_cache_ttl_secs300300DEFAULTSets the time-to-live (TTL) in seconds for cached query results.UInt64
query_tagDEFAULTSets the query tag for this session.String
quoted_ident_case_sensitive11GLOBALSet to 1 for case-sensitive treatment of quoted names, or 0 for case-insensitive.UInt64
random_function_seed00DEFAULTSeed for random function.UInt64
recluster_block_size92771293599277129359DEFAULTSets the maximum byte size of blocks for recluster.UInt64
recluster_timeout_secs4320043200DEFAULTSets the seconds that recluster final will be timeout.UInt64
replace_into_bloom_pruning_max_column_number44DEFAULTMax number of columns used by bloom pruning for replace-into statement.UInt64
replace_into_shuffle_strategy00DEFAULTChoose shuffle strategy: 0 for Block, 1 for Segment level.UInt64
s3_storage_classSTANDARDSTANDARDDEFAULTDefault S3 storage class. Values: "STANDARD", "INTELLIGENT_TIERING".String
sandbox_tenantDEFAULTInjects a custom 'sandbox_tenant' into this session. For testing purposes only.String
script_max_steps1000010000DEFAULTThe maximum steps allowed in a single execution of script.UInt64
short_sql_max_length20482048DEFAULTSets the maximum length for truncating SQL queries in short_sql function.UInt64
sort_spilling_batch_bytes2097152020971520DEFAULTSets the uncompressed size that merge sorter will spill to storage.UInt64
sort_spilling_memory_ratio6060LOCALSets the maximum memory ratio in bytes that a sorter can use before spilling data to storage during query execution.UInt64
spatial_runtime_filter_threshold10241024DEFAULTSets the maximum number of values in a spatial list for runtime filter generation.UInt64
spill_writer_memory_pool_size_mb2020DEFAULTSet the memory pool size (MB) for each spill writer.UInt64
spilling_file_formatparquetparquetDEFAULTSet the storage file format for spilling. Values: "arrow", "parquet".String
spilling_to_disk_vacuum_unknown_temp_dirs_limit1844674407370955161518446744073709551615DEFAULTSet the maximum number of directories to clean up for unexpected interrupted queries.UInt64
sql_dialectPostgreSQLPostgreSQLDEFAULTSets the SQL dialect. Available values: "PostgreSQL", "MySQL", "Experimental", "Hive", "Prql".String
statement_queue_ttl_in_seconds1515DEFAULTInterval in seconds between lease renewal operations with the meta service.UInt64
statement_queued_timeout_in_seconds00DEFAULTThe maximum waiting seconds in the queue. The default value is 0 (no limit).UInt64
storage_fetch_part_num22DEFAULTSets the number of partitions that are fetched in parallel from storage during query execution.UInt64
storage_io_max_page_bytes_for_read524288524288DEFAULTSets the maximum byte size of data pages that can be read from storage in a single I/O operation.UInt64
storage_io_min_bytes_for_seek4848DEFAULTSets the minimum byte size of data that must be read from storage in a single I/O operation when seeking a new location.UInt64
storage_read_buffer_size10485761048576DEFAULTSets the byte size of the buffer used for reading data into memory.UInt64
stream_consume_batch_size_hint00DEFAULTHint for batch size during stream consumption. Set it to 0 to disable it.UInt64
system_tables_count_db_concurrency1616DEFAULTSets the DB-level concurrency used by system.tables count optimization.UInt64
table_lock_expire_secs3030DEFAULTSets the seconds that the table lock will expire in.UInt64
timezoneUTCUTCDEFAULTSets the timezone.String
trace_sample_rate11DEFAULTSetting the trace sample rate. The value should be between '0' and '100'.UInt64
udf_cloud_import_presign_expire_secs259200259200DEFAULTPresign expiry for cloud UDF stage imports.UInt64
unquoted_ident_case_sensitive00DEFAULTSet to 1 to make unquoted names case-sensitive, or 0 for case-insensitive.UInt64
use_legacy_query_executor00DEFAULTFallback to legacy query executor.UInt64
use_parquet200DEFAULTThis setting is deprecated.UInt64
warehouseDEFAULTPlease use the USE WAREHOUSE statement to set the warehouse.String
week_start11DEFAULTSpecifies the first day of the week (used by week-related date functions).UInt64
window_num_partitions256256DEFAULTSets the number of partitions for window operator.UInt64
window_partition_sort_block_size6553665536DEFAULTSets the block size of data blocks to be sorted in window partition.UInt64
window_partition_spilling_memory_ratio6060GLOBALSets the maximum memory ratio in bytes that a window partitioner can use before spilling data to storage.UInt64
window_spill_unit_size_mb256256DEFAULTSets the spill unit size (MB) for window operator.UInt64