GUC Parameters ​

_INDEX_BUFFER_SIZE ​

Parameter Description
A global cache used to store read-only copies of index hot pages (currently mainly B-tree index root node pages). This cache is intended to reduce contention for high-concurrency access to index pages, thereby improving system concurrency performance.

Value Range

• Minimum value: 16K (16,384 bytes)

• Maximum value: 32T (35,184,372,088,832 bytes)

• Unit: byte

Default Value
8M (8,388,608 bytes)

_OPT_CBO_STAT_SAMPLING_LEVEL ​

Parameter Description
Sets the dynamic sampling level, which is used by the optimizer to collect statistics information. This parameter determines whether the optimizer adopts dynamic sampling and the sampling ratio when generating an execution plan in the absence of statistics information or when the statistics information is outdated.

Value Range

• Integer

• Minimum value: 0, indicating that the dynamic sampling feature is disabled.

• Maximum value: 9

Default Value
0

_PREFETCH_ROWS ​

Parameter Description
Sets the number of rows that the system prefetches into the cache during data retrieval.

Value Range

• Integer

• Minimum value: 1

• Maximum value: 2^32 - 1

Default Value
100

CHECKPOINT_PERIOD ​

Parameter Description
Sets the maximum period between two checkpoints executed by the system. When this preset period is reached, a checkpoint is automatically triggered.

Feature

• Alias: CHECKPOINT_TIMEOUT (modification is not recommended).

Value Range

• Integer, unit: second

• Minimum value: 1

• Maximum value: 2^32 - 1

Default Value
300

CHECKPOINT_PAGES ​

Parameter Description
Sets the threshold for the number of dirty pages that triggers a checkpoint. When the number of dirty pages in the database buffer reaches this value, a checkpoint is automatically triggered.

Feature

• Alias: CHECKPOINT_INTERVAL (modification is not recommended).

Value Range

• Integer, unit: page

• Minimum value: 1

• Maximum value: 2^32 - 1

Default Value
100,000

REACTOR_THREADS ​

Parameter Description
Sets the number of threads used for I/O listening. These threads are responsible for receiving and distributing connection requests and network packets from clients.

Configuration Recommendation

• Recommended value: OPTIMIZED_WORKER_THREADS / 50.

• Risk warning:

  1. Setting an excessively large value consumes additional CPU, memory, and thread resources. When system resources are insufficient, it may cause abnormal database operation.

  2. The value must not exceed the total number of Agent threads.

Value Range
[1, 10000]

Default Value
1

OPTIMIZED_WORKER_THREADS ​

Parameter Description
Sets the number of optimal worker threads for the database. This parameter determines whether the system adopts binding mode or separation mode between sessions and threads.

Working Mode

• When the number of sessions > OPTIMIZED_WORKER_THREADS, the system switches to the session-thread separation mode; otherwise, it uses the binding mode.

Relationship with MAX_WORKER_THREADS

1. At startup: If OPTIMIZED_WORKER_THREADS > MAX_WORKER_THREADS in the configuration file, MAX_WORKER_THREADS is automatically adjusted to the value of this parameter.

2. After startup:

   • It can be set online to any value less than or equal to the current MAX_WORKER_THREADS.

   • If an attempt is made to set it to a value greater than the current MAX_WORKER_THREADS, the system reports an error.

3. After restart: MAX_WORKER_THREADS retains its original value and is not changed by the modification of this parameter.

Configuration Recommendation and Constraints

• Upper limit for the maximum value: It is not recommended to exceed the value of the SESSIONS parameter; otherwise, thread resources will be left idle and wasted.

• Resource consumption:

  • Each worker thread occupies approximately 0.5 MB+ of memory.

  • Setting it too large will significantly increase CPU and memory overhead, and may cause database exceptions when resources are insufficient.

• Absolute upper limit: Because security functions are restricted to a maximum input memory of 2 GB, this parameter cannot exceed 10000.

Value Range
[2, 10000]

Default Value
100

MAX_WORKER_THREADS ​

Parameter Description
Sets the maximum capacity of the database worker thread pool.

Relationship with OPTIMIZED_WORKER_THREADS

1. Numeric relationship: The value of this parameter must be greater than or equal to OPTIMIZED_WORKER_THREADS. If an attempt is made to set it to a value smaller than the current OPTIMIZED_WORKER_THREADS, the system reports an error.

2. Thread expansion: When the number of active sessions exceeds OPTIMIZED_WORKER_THREADS, the number of worker threads can continue to expand from the optimal value until it reaches the maximum value set by this parameter. That is, the number of additionally expandable threads is MAX_WORKER_THREADS - OPTIMIZED_WORKER_THREADS.

Configuration Recommendation and Constraints

• Reference for the maximum value: OPTIMIZED_WORKER_THREADS is not recommended to exceed the value of the SESSIONS parameter; otherwise, thread resources will be wasted.

• Resource consumption:

  • Each worker thread occupies approximately 0.5 MB+ of memory.

  • Setting it too large will significantly increase CPU and thread management overhead, and may cause database exceptions when resources are insufficient.

• Absolute upper limit: Limited by the memory of system security functions, the maximum value of this parameter is 10000.

Value Range
[2, 10000]

Default Value
100

STATS_LEVEL ​

Parameter Description
Controls whether to collect and count the Data Manipulation Language (DML) operation information performed on tables.

Feature

• Alias: STATISTICS_LEVEL (modification is not recommended).

Value Range and Behavior

Default Value
TYPICAL

use_bison_parser ​

Parameter Description
Specifies whether to use the Bison parser to parse common SQL statements.

This parameter takes effect immediately after being set. For details about how to set it, see the corresponding setting method in Configuring Database System Parameters.

Value range: Boolean

• true indicates that the Bison parser is used.

• false indicates that the native parser is used.

• on and off can be used to set this parameter only when the Bison parser is enabled.

Default value: false

BUF_POOL_NUM ​

Parameter Description
Sets the number of partitions for the data buffer.

Value Range
[0, 1024]

Default Value
0

NOTE

• When this parameter is not configured or explicitly set to 0, the system automatically calculates and adjusts a reasonable number of partitions based on the actual value of DATA_BUFFER_SIZE.

BUFFER_PAGE_CLEAN_PERIOD ​

Parameter Description
Controls the execution interval of the Page Clean scheduled task (unit: second). Page Clean is a mechanism for quickly cleaning dirty pages in the buffer. In regular business scenarios, this scheduled task usually does not need to be enabled, because dirty page cleaning can be automatically triggered by the buffer eviction mechanism. However, in sustained high-load performance test scenarios such as TPC-C and sysbench, if the disk I/O capability is strong, this scheduled task can be enabled to accelerate dirty page flushing to disk.

Value range: integer, [0, 2^32 - 1]
Default value: 0 (Scheduled cleaning disabled)

INTERCONNECT_CHANNEL_NUM ​

Parameter Description
Configures the number of internal communication (MES) connection channels.

Value range: Integer, [1, 32]
Default value: 4

LOG_BUFFER_COUNT ​

Parameter Description
Sets the number of log buffers.

Value range: Integer, (0, 16]
Default value: 4

LOG_BUFFER_SIZE ​

Parameter Description
Defines the size of each log buffer, which is used to cache REDO logs. Properly increasing this value helps improve data write performance. Adjust it based on the actual available memory.

Value range: Integer, [1M, 128M] (Unit: byte)
Default value: 4M

TIMED_STATS (Alias: TIMED_STATISTICS, not recommended) ​

Parameter Description
Controls whether to collect time-related statistics.

• TRUE: Collects and stores time-related statistics (queryable through the DV_SYS_STATS dynamic view or trace files)

• FALSE: Sets all time-related statistics values to 0

Default value: TRUE

SQL_STAT ​

Parameter Description
Specifies Whether to enable SQL performance statistics views. The setting applies to the entire database instance.

• TRUE: Enables SQL performance statistics.

• FALSE: Disables SQL performance statistics.

Default value: TRUE

MES_ELAPSED_SWITCH ​

Parameter Description
Specifies whether to enable the latency statistics function for MES internal communication.

• TRUE: Enables latency statistics.

• FALSE: Disables latency statistics.

Default value: FALSE

SESSIONS ​

Parameter Description
Sets the maximum number of concurrent sessions in a database instance. The number of concurrent sessions at runtime must not exceed this value.

Session Composition ​

• System-reserved sessions:

  • Fixed 32 sessions: used for internal tasks (such as resource reclamation, checkpoint, etc.)

  • PARALLEL_MAX_THREADS sessions: used for the SQL parallel execution framework

• EMERG user sessions: the quantity is controlled by the SUPER_USER_RESERVED_SESSIONS parameter

• Regular user sessions:

  • At least 1 must be configured; otherwise, the database cannot start.

  • When parallel replay is disabled:

    Number of regular user sessions = SESSIONS − 32 − PARALLEL_MAX_THREADS − SUPER_USER_RESERVED_SESSIONS

  • When parallel replay is enabled (standby only):

    Available sessions = SESSIONS − 32 − PARALLEL_MAX_THREADS − SUPER_USER_RESERVED_SESSIONS − LOG_REPLAY_PROCESSES − REPLAY_PRELOAD_PROCESSES
    (The behavior on the primary is the same as when parallel replay is disabled.)

Configuration Recommendation ​

• Generally, it is recommended to set it to OPTIMIZED_WORKER_THREADS × 1.2

• On DN nodes, SESSIONS < OPTIMIZED_WORKER_THREADS is recommended.

Effective Mode ​

• Take effect immediately: The new value ≥ the initial value init_count and ≤ 1.5 × init_count

• Take effect after a restart: The new value < init_count or > 1.5 × init_count (but ≤ 16320)

  init_count is the SESSIONS value at database startup. The system reserves 50% of the expansion space by default for dynamic adjustment, so real-time adjustment supports only expansion, not reduction.

Precautions ​

1. After SUPER_USER_RESERVED_SESSIONS is modified, a restart is required for the change to take effect.

2. Before adjusting SESSIONS, ensure that _MAX_RM_COUNT satisfies the constraint conditions.

3. If SESSIONS > 13600, _MAX_RM_COUNT must be a multiple of 64 and greater than SESSIONS. It is recommended to set it to 16320.

Value range: integer Lower limit = 33 + PARALLEL_MAX_THREADS + SUPER_USER_RESERVED_SESSIONS Upper limit = 16320
Default value: 200

USE_RBP ​

Parameter Description
Master switch for RBP. After it is enabled, the database can work with RBPS to save and read the pages required for recovery, shortening the recovery time after an instance or node failure.

Disabled by default: When RBPS is not deployed or accelerated recovery is not required, disabling this capability avoids introducing additional resource overhead.

Value Range

• TRUE: Enables RBP.

• FALSE: Disables RBP.

Default Value
FALSE

Effective Mode
Takes effect after a restart.

Configuration Recommendation

• When RBPS is not deployed or only ordinary recovery is performed, keep the default value.

• When RBPS is required for accelerated recovery, set this parameter to TRUE for all related database nodes, and confirm that the RBPS service has been deployed and the network is reachable.

RBP_FOR_RECOVERY ​

Parameter Description
Controls whether RBP acceleration is allowed during the recovery phase. This parameter must be used together with USE_RBP=TRUE; when RBPS is available and the recovery conditions are met, the database can use pages in RBPS to accelerate the recovery.

Enabled by default: After you explicitly enable RBP, it is expected that the recovery phase directly uses the acceleration capability; this independent switch is also retained to facilitate temporarily disabling accelerated recovery.

Value Range

• TRUE: RBP is allowed during the recovery phase.

• FALSE: RBP is not used during the recovery phase.

Default Value
TRUE

Effective Mode
Takes effect after a restart.

Configuration Recommendation

• Keep the default value for regular accelerated recovery scenarios.

• When RBP accelerated recovery needs to be temporarily disabled, set it to FALSE and restart the instance.

RBP_IP ​

Parameter Description
Configures the target address list for the database to connect to RBPS.

In a two-node DTC deployment, the database side is usually configured to connect to the peer RBPS. It is recommended that RBP communication use a different NIC from service access.

Value Range

• A non-empty string.

• A single address or a comma-separated list of addresses can be configured.

Default Value
127.0.0.1

Effective Mode
Takes effect after a restart.

Configuration Recommendation

• In a two-node DTC scenario, the peer RBPS address is usually configured, for example, [ip1, ip2].

• If RBP uses a dedicated NIC, it is recommended to configure the address on the RBP network instead of the service access address.

RBP_PORT ​

Parameter Description
Configures the TCP port for connecting the database to RBPS. This port must be consistent with the listening port configured in the RBPS server rbps.conf.

Value Range
Integer, [1024, 65535]

Default Value
2611

Effective Mode
Takes effect after a restart.

Configuration Recommendation

• If the port needs to be changed, the RBP_PORT of all database nodes and the listening ports of all related RBPS servers must be modified synchronously, and the corresponding components need to be restarted.

LOCAL_RBP_HOST ​

Parameter Description
Configures the local address that the current database instance reports to RBPS. This parameter is used to identify the source database node in RBPS-side logs, status records, and read/write phase notifications. It does not determine which RBPS the database connects to; the connection target is determined by RBP_IP.

The default value is the local loopback address.

Value Range
A non-empty string.

Default Value
127.0.0.1

Effective Mode
Takes effect after a restart.

Configuration Recommendation

• In a cluster deployment, configure it as the RBP communication IP or stable hostname that can be recognized and distinguished by RBPS for the current database node.

• If RBP uses an independent NIC, it is recommended to configure the address corresponding to the independent NIC.

RBP_RT_ANALYSIS ​

Parameter Description
Controls whether to enable the DTC RBP runtime analysis capability. After it is enabled, the database prepares part of the information required for recovery in advance during runtime, thereby reducing the recovery preparation time after a fault occurs.

Disabled by default: Runtime analysis introduces additional resource overhead and should be enabled on demand in scenarios where reducing recovery time is explicitly required.

Value Range

• TRUE: Enables runtime analysis.

• FALSE: Disables runtime analysis.

Default Value
FALSE

Effective Mode
Takes effect after a restart.

Configuration Recommendation

• This parameter is meaningful only when USE_RBP=TRUE and RBP_FOR_RECOVERY=TRUE.

• When sensitive to runtime resource consumption, or when not in a DTC accelerated recovery scenario, keep the default value.

RBP_RT_PARSE_WORKERS ​

Parameter Description
Configures the number of parsing worker threads for DTC RBP runtime analysis.

Defaults to 2: In a two-node scenario, two parsing worker threads can balance analysis efficiency and CPU usage, avoiding excessive background overhead during initial deployment.

Value Range
Integer, [1, 8]

Default Value
2

Effective Mode
Takes effect after a restart.

Configuration Recommendation

• If the runtime analysis processing speed is insufficient and CPU resources are sufficient, the value can be increased appropriately.

• If CPU pressure is high or analysis pressure is low, you can retain the default value or reduce the value.

RBP_RT_PAGE_OWNER_WORKERS ​

Parameter Description
Configures the number of worker threads for page ownership processing in DTC RBP runtime analysis.

Defaults to 4: Page ownership processing usually requires a higher degree of concurrency than the parsing phase, and appropriately increasing the number of worker threads can reduce the risk of recovery information processing backlog.

Value Range
Integer, [1, 8]

Default Value
4

Effective Mode
Takes effect after a restart.

Configuration Recommendation

• Keep the default value in normal cases.

• If the runtime analysis backlog is significant and memory and CPU resources are sufficient, increase the value appropriately.

RBP_ASSEMBLE_MAX_SCAN ​

Parameter Description
Configures the maximum number of candidates to be scanned in a single pass when RBP writes pages. Increasing this value appropriately can improve batch page sending efficiency, but may also increase scanning overhead.

Defaults to 300: This value strikes a balance between batch sending efficiency and runtime overhead, and is suitable for most scenarios.

Value Range
Integer, [100, 1000000]

Default Value
300

Effective Mode
Takes effect immediately.

Configuration Recommendation

• Keep the default value in normal cases.

Database Parameter Configuration Description ​

1. Global Shared Memory ​

TEMP_BUFFER_SIZE ​

Function: Configures the total size of the temporary buffer, which is used to support temporary operations such as materialized table creation and data sorting.

• At system startup, the buffer is evenly divided into multiple temporary pools based on TEMP_BUFFER_SIZE (total size) and TEMP_POOL_NUM (number of temporary pools).

• The minimum size of a single temporary pool is 2 GB. If TEMP_BUFFER_SIZE < 2 GB, only one temporary pool is created.

• Supports online dynamic expansion:

  • The space expanded each time must be greater than or equal to temp_buffer_extend_threshold.

  • Number of newly added temporary pools = MIN(expanded value / size of a single temporary pool, 1)

  • The maximum total number of temporary pools is 128.

• Online reduction is not supported: To reduce this parameter, modify the database configuration file or use SCOPE=PFILE, and restart the database for the change to take effect.

Value range: integer, [32M, 4T] (Unit: byte)
Default value: 32M
Installation script default value: When install.py is used for installation, it is automatically set to 1G.

DATA_BUFFER_SIZE ​

Function: Sets the size of the data buffer, which is used to cache recently accessed data pages and improve data access efficiency.

• It should be configured reasonably based on the physical memory of the server.

• Takes effect dynamically, but is subject to the following parameter constraints:

  • Minimum value = BUFFER_UINT_SIZE

  • Maximum value = BUFFER_UINT_SIZE × BUF_POOL_NUM × 1024

• If BUFFER_CHECK_MEMORY = TRUE, the configured value must not exceed the total physical memory, otherwise an error is triggered in the following scenarios:

  • Parameter configuration phase

  • Online expansion operation

  • Loading configuration at database startup

Value range: integer, [64M, 32T], supported units: B / K / M / G / T (Byte by default when no unit is specified)
Default value: 128M
Configuration recommendation:

• It is recommended to set DATA_BUFFER_SIZE to an integer multiple of BUFFER_UINT_SIZE × BUF_POOL_NUM to reduce memory fragmentation.

• Only online expansion is supported; to decrease it, modify the configuration file or use SCOPE=PFILE, and restart the database for it to take effect.

Installation script default value: When install.py is used for installation, it is automatically set to 2G.

SHARED_POOL_SIZE ​

Function: Configures the total size of the shared pool, including shared memory structures such as the Lock Pool, SQL Pool, and DC Pool.

• The minimum value is affected by init_lock_pool_pages:

  • Only when init_lock_pool_pages = 128 is configured, SHARED_POOL_SIZE is allowed to be set to the minimum value of 68M.

• The SHARED_POOL_SIZE of the standby node must be greater than or equal to the primary node value; otherwise, the 68M minimum configuration cannot be used.

• After increasing this parameter, it is recommended to evaluate whether init_lock_pool_pages needs to be increased accordingly.

• Dynamic adjustment restrictions:

  • Only online expansion is supported; online reduction is not supported.

  • The single expansion amount (expand_size) must be greater than or equal to MIN(512M, current SHARED_POOL_SIZE); otherwise, it takes effect only after a restart.

  • A maximum of 15 dynamic expansions is supported; beyond this limit, subsequent modifications take effect only after the database is restarted.

Value range: integer, [68M, 32T] (Unit: byte)
Default value: 128M
Configuration recommendation:

• To reduce the value, you must modify the configuration file or use SCOPE=PFILE, and restart the database for the change to take effect.

Installation script default value: When install.py is used for installation, it is automatically set to 1G.

2. Checkpoint ​

CHECKPOINT_IO_CAPACITY ​

Function: Controls the number of dirty pages (in pages) written to the disk each time a checkpoint is triggered.

Value range: Integer, [1, 4096]
Default value: 1024

_CHECKPOINT_TIMED_TASK_DELAY ​

Function: Sets the interval between two scheduled checkpoint tasks (in milliseconds).

• Scheduled checkpoint is jointly controlled by checkpoint_pages and checkpoint_period. This parameter is used to adjust the execution frequency.

Value range: integer, [0, 2^32 - 1] (unit: millisecond)
Default value: 100

3. Transaction ​

_UNDO_AUTO_SHRINK ​

Function: Controls whether to enable the automatic shrink feature of Undo Segment.

Value range:

• true: Enables automatic shrink.

• false: Disables automatic shrink.

Default value: true

4. Process ​

DBWR_PROCESSES ​

Function: Sets the number of background dirty page writer (DBWR) threads.

• Increasing this value improves I/O concurrency, but increases CPU and memory overhead.

• The default configuration varies by installation package type:

  • Shared storage installation package: defaults to 36

  • Other types of installation packages: defaults to 8

• Special recommendation: When FILE_OPTIONS = ASYNCH or SETALL (that is, asynchronous I/O is enabled), it is recommended to set this parameter to 1, because asynchronous I/O already provides high concurrency capability and does not require additional DBWR threads.

Value range: integer, [1, 36]
Default value: 1

5. Session-level Parameters ​

_DOUBLEWRITE ​

Function: Whether to enable the double write mechanism, which improves the reliability of data writes and prevents partial write issues.

Value range:

• TRUE: Enables double write.

• FALSE: Disables double write.

Default value: TRUE

Note: For all parameters that only support online expansion, if a reduction is required, it must be modified through the configuration file or SCOPE=PFILE, and takes effect only after restarting the database instance.