Columnar engine flags
Stay organized with collections
Save and categorize content based on your preferences.
This page describes the database flags that AlloyDB for PostgreSQL uses to enable and
manage features specific to its columnar
engine. For a list of all database flags
that AlloyDB supports, see Supported database
flags.
Flags marked with Instance restarts mean that AlloyDB
restarts an instance whenever you set, remove, or modify this flag on that
instance. The flag's value persists for the instance until you modify it again.
Specifies a list of columns for AlloyDB to add to your cluster's
column store, apart from auto-columnarization. These columns can come from any
tables within your cluster.
The flag's value is formatted as a comma-separated list of items. Each item
specifies one or more columns from a single table, expressed in this format:
Specifies the frequency that AlloyDB runs its
auto-columnarization process on the
instance.
The value is a string in this format:
'EVERY NUMBERTIME_UNITS'
Replace the following:
NUMBER: A positive integer.
TIME_UNITS: Either DAYS or HOURS.
If not defined, then AlloyDB runs the auto-columnarization
process once every hour.
Sizing and query processing
google_columnar_engine.memory_size_in_mb
Type Integer
Valid values From 128 to 70% of instance memory
Default 30% of instance memory
Instance restarts Yes
The amount of instance memory, expressed in mebibytes (MiB), that
AlloyDB allocates to the column store.
If this flag is not set, then AlloyDB allocates
30% of the instance's memory to the column store.
google_columnar_engine.enable_columnar_scan
Type Boolean
Default on
Instance restarts No
If on, allows AlloyDB to access data in the column store when
handling queries.
If this flag is off while google_columnar_engine.enabled is
on, then the columnar engine stays enabled and retains its column store, but
AlloyDB does not use columnar data when handling queries.
google_columnar_engine.enable_index_caching
Type Boolean
Default off
Instance restarts No, if the columnar engine is already enabled, that is when google_columnar_engine.enabled=on.
Indicates if you can load ScaNN indexes into the columnar engine to accelerate vector search.
Specifies the column store's data-refresh threshold, expressed as a percentage.
AlloyDB refreshes the data within a columnar unit—a set of data
blocks within a columnar table—when all of these conditions are met:
The columnar unit's table has exceeded the scan-threshold defined by the
separate refresh_threshold_scan_count
flag.
The percentage of the columnar unit's underlying data blocks that have
become invalid exceeds the value of the refresh_threshold_percentage
flag. Columnar data blocks become invalid due to changes in the row-based
data that the columnar table draws upon.
AlloyDB keeps a running count of the table scans performed on
each columnar table since that table's most recent DML query. The value of the
refresh_threshold_scan_count flag specifies the minimum value of this counter
on any columnar table before AlloyDB can refresh its data. From
there, AlloyDB uses the invalid-block threshold defined by
refresh_threshold_percentage to determine
when to refresh the table's data.
If this flag's value is set to 0, then AlloyDB
disregards table scans to determine when to refresh columnar data, using only
refresh_threshold_percentage instead.
google_job_scheduler.maintenance_cpu_percentage
Type Integer
Valid values 0 ... 100
Default 20
Instance restarts No
The maximum percentage of an instance's CPU capacity that the columnar engine
can apply to background processing.
google_job_scheduler.max_parallel_workers_per_job
Type Integer
Valid values 0 ... 262143
Default 2
Instance restarts No
The maximum number of parallel workers that the columnar engine can create to
handle any single background processing operation.
Vectorized join
google_columnar_engine.enable_vectorized_join
Type Boolean
Default off
Restarts instance No
If on, then AlloyDB automatically uses multithreaded vectorized
operations to improve the performance of queries that join multiple tables.
[[["Easy to understand","easyToUnderstand","thumb-up"],["Solved my problem","solvedMyProblem","thumb-up"],["Other","otherUp","thumb-up"]],[["Hard to understand","hardToUnderstand","thumb-down"],["Incorrect information or sample code","incorrectInformationOrSampleCode","thumb-down"],["Missing the information/samples I need","missingTheInformationSamplesINeed","thumb-down"],["Other","otherDown","thumb-down"]],["Last updated 2025年11月12日 UTC."],[],[]]