[engine] Settings
Engine settings in vectorwise.conf affect the query execution engine.
enable_aggregation_disk_spilling
Enables or disables spilling to disk for hash aggregations. Enabling this option prevents out-of-memory conditions from occurring when aggregate results are too large to fit into memory. Valid values are true and false.
Default: false
enable_hashjoin_disk_spilling
Enables or disables spilling to disk for hash joins. Enabling this option prevents out-of-memory conditions from occurring when the hash table built from one of the join inputs is too large to fit into memory. Valid values are true and false.
Default: false
enable_reuse_disk_spilling
Enables or disables spilling to disk for reused repeating query plan parts. Enabling this option prevents out-of-memory conditions from occurring when the buffered output of a reused query part is too large to fit into memory. Valid values are true and false.
Default: false
listagg_group_maxlen
Specifies the maximum result length of the LISTAGG() aggregate function in ASCII bytes.
Limits: 1 to 32000
Default: 4000
max_parallelism_level
Defines the maximum number of threads used by a single query. This is a hard limit.
A value of 1 disables parallel query execution. A value larger than 1 enables parallel query execution.
Max_parallelism_level is related to the num_cores parameter. The execution engine tries not to use more than the max_parallelism_level number of cores at a time for any query.
Limits: 0 to 256
For VectorH, max_parallelism_level is set on the master node only, but governs the total level of parallelism throughout the cluster. A good value is the total number of cores across all nodes.
Default: Total number of cores in the cluster.
update_parallelism_level
Defines the forced level of parallelism used by update (DML) queries. This setting can be used to improve performance of update queries that can benefit from parallelization. Values:
-1
(Default) Treats update queries as normal queries, load balances the number of cores granted to them using the usual algorithms using max_parallelism_level (or max_update_parallelism_level).
0
Uses max_parallelism_level (or max_update_parallelism_level, if set) and lets the load balancing algorithm establish the target number of cores for update queries in the usual way. Protects the cores granted to update queries by excluding them from the core pool used for load balancing. Same as max_update_parallelism_level >= 0.
level
Uses the specified hard-coded parallelism level (must be greater than 0) for update queries that will be granted irrespectively of load balancing. Protects the cores granted to update queries by excluding them from the core pool used for load balancing. Takes precedence over max_parallelism_level or max_update_parallelism_level.