Standby Server

hot_standby

Parameter description: Specifies whether the standby node is allowed to accept connections and queries after it is restored to the minrecovery point.

This is a POSTMASTER parameter. Set it based on instructions in Table 1.

• If this parameter is set to on, wal_level must be set to hot_standby or higher. Otherwise, the database startup fails.
• In a distributed system, hot_standby cannot be set to off, because this setting can affect other features of the HA system.
• If the hot_standby parameter was disabled and the wal_level parameter was set to a value smaller than that of hot_standby, perform the following operations to ensure that the logs to be replayed on the standby node can be queried on the standby node before enabling the hot_standby parameter again:
  1. Change the value of wal_level of the primary and standby nodes to the value of hot_standby or a higher value, and restart the instances for the change to take effect.
  2. Perform the checkpoint operation on the primary node and query the pg_stat_get_wal_senders() function to ensure that the value of receiver_replay_location of each standby node is the same as that of sender_flush_location of the primary node. Ensure that the value adjustment of wal_level is synchronized to the standby nodes and takes effect, and the standby nodes do not need to replay low-level logs.
  3. Set the hot_standby parameter of the primary and standby nodes to on, and restart the instances for the setting to take effect.
• If the read on standby function is enabled, conflicts between replay and query on the standby node may cause the query to cancel. The error information is as follows:
  • ERROR: canceling statement due to conflict with recovery
  • ERROR: terminating connection due to conflict with recovery
• During the standby node read in serial and parallel replay modes, when the primary node rebuilds indexes online, an error may be reported when the standby node is read. Try again later. Error information:
  • could not open relation with OID xxx during recovery delete object, please try again later
  • Catalog is missing xxx attribute(s) for relid xxx
  • cache lookup failed for index xxx, refilenode:xxx, name:"xxx"
  • could not find pg_class entry for xxx
• During the standby node read in the case of ultimate RTO, the VM with small memory configuration may report an error indicating insufficient memory. You can enable the GUC parameter exrto_standby_read_opt (enabled by default) to effectively reduce the memory and I/O overhead.
• The standby node does not support autonomous transactions.
• Use sysbench to test the read performance of the standby node in typical scenarios: The primary node executes 100 concurrent update services. The primary and standby nodes both execute 200 concurrent read services at the same time. When the I/O and CPU are not limited, the read performance of the standby node in serial replay mode is not lower than 80% of that of the primary node, and the read performance of the standby node in ultimate RTO deteriorates by no more than 10% compared with that of the standby node in serial replay mode.

Value range: Boolean

• on: allowed.
• off: not allowed.

Default value: on

max_standby_archive_delay

Parameter description: Specifies the wait period before queries on a standby node are canceled when the queries conflict with WAL processing and archiving in hot standby mode. In the current version, the setting does not take effect and is controlled by the max_standby_streaming_delay parameter.

Parameter type: integer.

Unit: millisecond

Value range: –1 to 2147483647. The value –1 indicates that the standby node waits until the conflicting queries are complete.

Default value: 3s (that is, 3000 ms)

Setting method: This is a SIGHUP parameter. Set it based on instructions in Table 1.

Setting suggestion: Retain the default value.

max_standby_streaming_delay

Parameter description: Specifies the wait period before queries on a standby node are canceled when the queries conflict with WAL data receiving through streaming replication in hot standby mode. If this parameter is set to a large value or the service load is heavy, an error may be reported for waiting for transaction replay and flushing to disks.

This is a SIGHUP parameter. Set it based on instructions in Table 1.

Value range: an integer ranging from –1 to 2147483647. The unit is ms. ‬

–1 indicates that the standby node waits until the conflicting queries are complete. In the scenario where serial or parallel replay is enabled, if the system detects that the query thread and replayer thread are in the deadlock state, the system still cancels the query to prevent replay blocking.

Default value: 3s (that is, 3000 ms)

wal_receiver_status_interval

Parameter description: Specifies the maximum interval for notifying the primary node of the WAL receiver status.

This is a SIGHUP parameter. Set it based on instructions in Table 1.

Value range: an integer ranging from 0 to INT_MAX/1000. The unit is s.

If this parameter is set to 0, the standby node does not send information, such as the log receiving location, to the primary node. As a result, the transaction commit on the primary node may be blocked, and the switchover may fail. In normal service scenarios, you are advised not to set this parameter to 0.

Default value: 5s

hot_standby_feedback

Parameter description: Specifies whether the ID of the oldest transaction active on standby nodes will be sent to the primary node, preventing a query conflict.

This is a SIGHUP parameter. Set it based on instructions in Table 1.

Value range: Boolean

• on indicates that the standby node is allowed to send the result of a query performed on it to the primary node.
• off indicates that the standby node is not allowed to send the result of a query performed on it to the primary node.

Default value: off

wal_receiver_timeout

Parameter description: Specifies the maximum wait period for a standby node to receive data from the primary node.

This is a SIGHUP parameter. Set it based on instructions in Table 1.

Value range: an integer ranging from 0 to 2147483647. The unit is ms.

Default value: 6000ms (that is, 6s)

wal_receiver_connect_timeout

Parameter description: Specifies the timeout interval for a standby node to connect to the primary node.

This is a SIGHUP parameter. Set it based on instructions in Table 1.

Value range: an integer ranging from 0 to 2147483. The unit is s.

Default value: 2s

wal_receiver_connect_retries

Parameter description: Specifies the maximum attempts that a standby node connects to the primary node.

This is a SIGHUP parameter. Set it based on instructions in Table 1.

Value range: an integer ranging from 1 to 2147483647.

Default value: 1

wal_receiver_buffer_size

Parameter description: Specifies the memory buffer size for the standby nodes to store the received Xlog files.

This is a POSTMASTER parameter. Set it based on instructions in Table 1.

Value range: an integer ranging from 4096 to 1047552. The unit is KB.

Default value: 64MB (that is, 65536 KB)

primary_slotname

Parameter description: Specifies the slot name of the primary node corresponding to a standby node. This parameter is used for the mechanisms to verify the primary-standby relationship and delete WALs.

This is a SIGHUP parameter. Set it based on instructions in Table 1.

Value range: a string.

Default value: an empty string

enable_redo_atomic_operation

Parameter description: Specifies whether to use atomic operations or spinlocks to update the LSN of the current thread when parallel replay is enabled.

This is a POSTMASTER parameter. Set it based on instructions in Table 1.

Value range: Boolean

• on indicates that atomic operations are used for update.
• off indicates that spinlocks are used for update.

Default value: on

max_standby_base_page_size

Parameter description: Specifies the maximum storage space of base page files on the standby node after the ultimate RTO function is enabled.

This is a SIGHUP parameter. Set it based on instructions in Table 1.

Value range: a long integer ranging from 1048576 to 562949953421311. The unit is KB.

Default value: 268435456‬ (256 GB)

max_standby_lsn_info_size

Parameter description: Specifies the maximum size of LSN info files on the standby node after the ultimate RTO function is enabled.

This is a SIGHUP parameter. Set it based on instructions in Table 1.

Value range: a long integer ranging from 1048576 to 562949953421311. The unit is KB.

Default value: 268435456 (that is, 256 GB)

max_keep_csn_info_size

Parameter description: Specifies the maximum size of CSN info linked lists allowed by each DN on the standby node after the ultimate RTO function is enabled.

This is a SIGHUP parameter. Set it based on instructions in Table 1.

Value range: a long integer ranging from 16384 to 131072. The unit is KB.

Default value: 16384‬ (that is, 16 MB)

base_page_saved_interval

Parameter description: Specifies the interval for generating base pages on the standby node after the ultimate RTO function is enabled. For the same page, a base page is generated each time the value of this parameter is replayed.

This is a POSTMASTER parameter. Set it based on instructions in Table 1.

Value range: an integer ranging from 5 to 2000.

Default value: 400

standby_force_recycle_ratio

Parameter description: Specifies the percentage of files read by the standby node to trigger forcible recycling after the ultimate RTO function is enabled. When the total size of base page files exceeds the value of max_standby_base_page_size x standby_force_recycle_ratio or the total size of LSN info files exceeds the value of max_standby_lsn_info_size x standby_force_recycle_ratio, forcible recycling is triggered and some queries are canceled. When the value of standby_force_recycle_ratio is 0, forcible recycling is not started, and the setting of max_standby_base_page_size and max_standby_lsn_info_size does not take effect.

This is a SIGHUP parameter. Set it based on instructions in Table 1.

Value range: a double-precision floating-point number ranging from 0.0 to 1.0.

Default value: 0.8

standby_recycle_interval

Parameter description: Specifies the interval for the standby node to recycle read files after the ultimate RTO function is enabled. The thread for recycling read resources on the standby node attempts to clear read files on the standby node at the interval specified by this parameter.

This is a SIGHUP parameter. Set it based on instructions in Table 1.

Value range: an integer ranging from 0 to 86400‬. The unit is s.

Default value: 10

standby_max_query_time

Parameter description: Specifies the maximum query time supported on the standby node after the ultimate RTO function is enabled. If the query time exceeds the value of this parameter, the query will be canceled. Note: The time when the query is canceled is affected by the interval parameter standby_recycle_interval of the recycling thread and the time when the snapshot is obtained. Therefore, the actual execution time of the query on the standby node must be greater than the value of this parameter.

This is a SIGHUP parameter. Set it based on instructions in Table 1.

Value range: an integer ranging from 0 to 86400‬. The unit is s.

Default value: 600

exrto_standby_read_opt

Parameter description: Specifies whether to support read optimization of the standby node with ultimate RTO. This parameter is enabled by default. This parameter is not synchronized between the primary and standby nodes.

This is a POSTMASTER parameter. Set it based on instructions in Table 1.

Value range: Boolean. on indicates that the optimization is enabled, and off indicates that the optimization is disabled.

Default value: on

walrcv_writer_crc_check_level

Parameter description: Specifies whether to enable Xlog verification on the standby node in the primary/standby clusters in the streaming DR scenario. By default, the parameter is enabled only in the DR cluster.

Parameter type: integer.

Unit: none

Default value: 1

Setting method: This is a SIGHUP parameter. Set it based on instructions in Table 1.