|
Version
|
Description
|
|
2.25.11
|
-
New features
-
Bug fixes
-
Fixed an issue in HTAP scenarios where incompatible SQL statements were not rerouted to the transactional processing (TP) node.
-
Fixed an issue where load imbalances occasionally occurred after a read-only transaction was executed in a single session.
-
Fixed an issue where connections were dropped due to high transient latency on the primary node in a high availability architecture.
|
|
2.25.9
|
|
|
2.25.8
|
|
|
2.25.7
|
Bug fixes
|
|
2.25.5
|
-
New features
-
Bug fixes
-
Fixed an issue where new connections could be blocked during a fast in-place restart of the primary MySQL node after a crash.
-
Fixed an issue that could cause load imbalances for some persistent connections when proxy parameters or nodes were changed.
|
|
2.25.4
|
New features
-
Added a load balancing strategy based on the number of active requests. For more information, see Load balancing algorithms.
-
You can now configure the primary node as a fallback node for a read-only proxy endpoint.
|
|
2.25.3
|
Bug fixes
|
|
2.25.2
|
|
|
2.25.1
|
New features
|
|
2.24.3
|
Bug fixes
-
Fixed an issue where a SELECT ... FOR UPDATE statement failed if a column name was the keyword 'disable'.
-
Fixed an issue where sessions were unexpectedly disconnected due to wait_timeout.
-
Fixed a schema error in the transaction-level connection pool that occurred when a read-only node was added.
|
|
2.24.1
|
Bug fixes
-
Fixed an out-of-memory (OOM) issue caused by an excessively large recv-q for backend connections.
-
Fixed an issue where the schema for the transaction-level connection pool was incorrectly reused when a node was added.
|
|
2.9.5
|
|
|
2.9.3
|
Bug fixes
-
Fixed an issue that could cause load imbalances in read/write splitting mode for some persistent connection scenarios.
-
Fixed a memory leak issue introduced in version 2.9.1 that was caused by SHOW WHERE statements.
|
|
2.9.1
|
-
New features
-
Added support for configuring connection pool and latency threshold settings for proxy endpoints with read-only attributes.
-
Added support for persistent connections in failover scenarios.
-
Optimized persistent connections to support transactions without write operations at the RC isolation level.
-
Bug fixes
-
Fixed the thread id already exists error caused by thread ID rotation by minimizing rotation.
-
Fixed an issue where a KILL QUERY operation on a read-only endpoint might not be broadcast, preventing the request from being killed.
|
|
2.8.41
|
New features
Upgraded the kernel to community version 2.3.15.
|
|
1.14.8/1.14.9
|
Bug fixes
-
Fixed an issue where a KILL statement could cause a session to hang.
-
Fixed an issue where sessions on a read-only node with a weight of 0 could not be successfully killed.
-
Fixed an issue where single-connection bandwidth was limited in high-latency networks.
-
Fixed an issue where loads could become unbalanced after a faulty node recovered.
-
Fixed an issue where a task flow failed because the MHD thread became unresponsive.
-
Fixed a memory leak that occurred when handling short-lived connections.
-
Fixed an issue where the force_node syntax did not take effect.
|
|
1.14.7
|
|
|
1.14.6
|
New features
Added support for persistent connections during manual switchovers.
|
|
1.14.5
|
Bug fixes
Fixed an issue that could cause client disconnections after the connection pool feature was enabled by ensuring compatibility with the session tracker in MySQL 8.0.33.
|
|
1.14.3/1.14.4
|
Bug fixes
-
Fixed an issue where the proxy endpoint became unavailable because the dual-write prevention feature incorrectly identified the primary node. Error codes have been added for this feature, so clients now receive a specific error when it triggers.
-
Fixed an issue where authentication for a proxy endpoint failed if the account host was in CIDR format.
-
Fixed an issue where the first read request within a transaction was forwarded to a read-only node when transaction splitting was disabled.
|
|
1.14.1/1.14.2
|
New features
Added the dual-write prevention feature to avoid data inconsistencies caused by a dual-primary scenario during a failover. When this feature is enabled, the proxy endpoint becomes unavailable if multiple primary nodes exist.
|
|
1.13.42
|
Bug fixes
|
|
1.13.41
|
-
Performance improvements
-
Improved database proxy performance during high volumes of failed authentications.
-
Optimized the account authentication feature. In the event of incorrect account authentication, throttling is now applied at the account level instead of globally.
-
Bug fixes
-
Fixed an issue where transaction status was lost during transaction splitting.
-
Fixed an issue where environment variables were not reset in a transaction-level connection pool.
-
Fixed an issue where a prepared statement failed to close in a distributed transaction (XA transaction).
-
Fixed several internal exceptions.
|
|
1.13.33
|
|
|
1.13.27
|
|
|
1.13.25
|
Bug fixes
|
|
1.13.22
|
-
New features
-
Added support for using the SELECT LAST_INSERT_ID() statement to obtain the unique ID of a sharded table after an INSERT operation.
-
Added support for the FOUND_ROWS function.
-
Added support for routing COM_STATISTICS to a read-only instance.
-
Optimized the transaction-level connection pool.
-
Added support for routing GEO functions to a read-only instance.
-
Added several internal monitoring metrics.
-
Bug fixes
-
Connections are no longer sent to locked instances in ApsaraDB RDS.
-
Fixed routing errors caused by incorrect parsing of some statements.
-
Fixed an issue where stmt_exec failed in some scenarios.
-
Fixed an issue where the load data infile statement failed.
|
|
1.13.17
|
Bug fixes
Fixed several internal errors.
|
|
1.13.5
|
-
New features
-
Added support for routing all requests to a specific node (Force Node Connection). For example, the statement /*force_proxy_internal*/set force_node = 'pi-123'; routes all subsequent requests on the current connection to the specified node pi-123.
Catatan
If the specified node is abnormal, the proxy returns the error set force node 'pi-123' is not found, please check..
-
Added support for routing a single request to a specific node (Force Node Query). For example, the statement /*force_node='pi-123'*/ show processlist; routes only this request to the specified node pi-123.
Catatan
If the specified node is abnormal, the proxy returns the error 'force hint server node is not found, please check'..
-
Added several internal monitoring metrics.
-
Bug fixes
|
|
1.12.10
|
|
|
1.12.7
|
-
New features
-
Bug fixes
-
Fixed an issue that caused errors when running show processlist on a MySQL 8.0 instance.
-
Fixed several issues related to the transaction-level connection pool.
-
Fixed several issues that caused connection failures.
|
|
1.11.12
|
|
|
1.10.7
|
Bug fixes
Fixed several issues related to the session-level connection pool.
|
|
1.9.23
|
-
New features
-
Bug fixes
-
Fixed an issue where change user failed.
-
Fixed an issue where load file failed.
-
Fixed an issue where clients received an Exception: Packets out of order error due to an incorrect packet sequence.
-
Fixed an issue where connections to a read-only instance were dropped when the primary instance became abnormal.
|
|
1.9.14
|
|