PolarDB:PolarDB for MySQL 8.0.1

Category

Description

Release date

New feature and feature optimization

• The dbms_oss.delete_table_flie built-in stored procedure can be used to delete objects that are not needed in OSS.

• The database and table restoration feature is optimized. For example, databases and tables can be restored more quickly, and failed tasks can be rolled back.

2023-10-31

Fixed issue

• The following issue is fixed: Tables that contain IMCIs cannot be restored by point in time.

• The following issue is fixed: Excessive memory consumption by the GroupJoin operator can lead to query failures in some scenarios.

• The following issue is fixed: The row IDs that are generated after the checkpoint fail to be rolled back when the checkpoint is loaded.

• The following issue is fixed: When read-only nodes of X-Engine are accessed, the nodes cannot report information about version reference, resulting in space reclamation failures.

• The following issue is fixed: When a join operation is performed between an unsigned column and a signed column, the optimizer cannot accurately estimate the join selection rate.

• The following issue is fixed: Some queries that contain OR predicates cannot be executed because the query plan cannot be bound to the correct columns.

• The following issue is fixed: The database may be restarted when a query is terminated by the scheduler.

• The following issue is fixed: The I/O performance of IMCIs is degraded in deferred materialization scenarios.

• The following issue is fixed: In rare cases, a Multi-master Cluster Edition cluster crashes because the metadata lock of scheduled tasks is not released at the earliest opportunity.

• The following issue is fixed: When X-Engine tables are accessed, query execution becomes slower because indexes cannot be used to overwrite scan results after primary key indexes are accessed.

• The following issue is fixed: When an ApsaraDB RDS for MySQL instance is upgraded to a PolarDB for MySQL cluster, the migration link is interrupted due to the restart of the cluster.

• The following issue is fixed: When X-Engine tables or indexes are deleted, unexpected redo logs are generated due to the concurrency relationship between the deletion operation and an occasional SwitchMemtable operation in the background and therefore the replica node crashes.

Category

Description

Release date

New feature and feature optimization

• The rate at which read-only nodes read redo logs in large blob scenarios is improved.

• A stored procedure name in uppercase is supported when a built-in stored procedure is called.

2023-09-21

Fixed issue

• The following issue is fixed: Nodes crash due to backticks in the value of an input parameter when some built-in stored procedures are called.

• The following issue is fixed: Nodes may crash if only keywords are specified but not the table schema and table name in the last rule when multiple CCL rules are configured to delete the same type of SQL statements.

• The following issue is fixed: When a new read-only node is connected to the read/write node, if file extension logs exist after the checkpointing operation is performed on the read/write node, the read-only node crashes in rare cases during the upgrade of the node to the read/write node.

Category

Description

Release date

New feature and feature optimization

• The speed of archiving cold data in the CSV format to OSS is improved.

• The cost model of accessing non-covered indexes for equality condition queries is optimized.

• The opt_non_covering_key_cost parameter is added to specify whether to consider the initial positioning cost of secondary indexes.

• The worst_seek_records_threshold parameter is added to specify whether to limit the maximum I/O cost of accessing non-covered indexes for equality condition queries during cost estimation.

• The timeout degradation mechanism is supported for the semi-synchronous replication of redo logs.

2023-08-22

Fixed issue

• The following issue is fixed: When a CTE contains nested CTEs and is referenced multiple times in a single query, the JPPD optimization feature does not work as expected.

• The following issue is fixed: When the primary node of a cluster is stopped and then restarted, the secondary nodes of the cluster cannot be reconnected to the primary node.

• The following issue is fixed: An error may occur when the JSON_REMOVE function is used in the syntax for updating BLOB data.

• The following issue is fixed: Nodes may be unexpectedly restarted during transaction deadlock checks due to illegal memory access.

Category

Description

Release date

Fixed issue

• The following issue is fixed: The execution of DDL operations becomes slower when read-only nodes do not return the ACK response to the read/write node at the earliest opportunity.

• The following issue is fixed: When DROP PARTITION operations are performed on an X-Engine Edition cluster, all nodes in the cluster crash unexpectedly.

2023-08-17

Category

Description

Release date

Fixed issue

The following issue is fixed: A large number of CPU resources are occupied by a cluster of low specifications.

2023-07-13

Category

Description

Release date

New feature and feature optimization

• The SELECT INTO DEFAULT OSSOUTFILE statement is added.

• The innodb_polar_import_tablespace_strict_mode parameter is added. The default value of this parameter is ON, which indicates that when a table is restored, an error is returned if the table schema contains full-text indexes.

• IMCIs can be used to query the data of a specified partition in a partitioned table.

• The SQL Trace feature can be used to sample specific SQL statements.

2023-07-05

Fixed issue

• The following issue is fixed: Regular users can delete the sys account.

• The following issue is fixed: After a HA switchover between the primary and read-only nodes is performed, the read-only nodes are unexpectedly restarted when a partitioned table is restored.

• The following issue is fixed: A read-only node cannot be switched to the primary node when a multi-master cluster is being restored by point in time.

• The following issue is fixed: The stored procedures, custom functions, scheduled tasks, views, and triggers of a multi-master cluster are not isolated among different nodes.

• The following issue is fixed: The stored procedures, custom functions, scheduled tasks, views, and triggers of a multi-master cluster cannot be switched among different nodes.

• The following issue is fixed: After the query cache feature is enabled for a multi-master cluster, updates to cluster data cannot be synchronized to the Information_schema database.

• The following issue is fixed: When a large number of tables exist in a multi-master cluster, data clearing can be slow.

Category

Description

Release date

New feature and feature optimization

• IMCIs can be created on virtual columns.

• The recycle bin feature is supported on tables for which IMCIs are created.

• Hybrid partitioned tables can be created. For more information, see Create a hybrid partitioned table.

• The federated query engine allows the LIMIT or OFFSET clause to be pushed down to the remote database for execution.

• The federated query engine allows all compatible conditions to be pushed down to the remote server for execution and returns only the columns required in the query.

• The call dbms_dts.show_parallel_ranges (<schema_name>, <table_name>, [<suggestion_ranges>]) function is added to display the parallel ranges for data sharding when a primary key is added to a table.

• The cold data archiving feature is supported for tables that contain auto-increment primary keys.

• The query capabilities of OSS are optimized. If the multi-node elastic parallel query (ePQ) feature is enabled, a parallel query is executed when the value of the loose_csv_max_oss_threads system parameter is greater than 1.

• The error message that indicates an invalid format for an OSS foreign table file is optimized.

• The without validation option can be added to the ALTER TABLE PARTITION BY RANGE statement to convert non-partitioned tables to range partitioned tables.

• The ALTER TABLE REMOVE PARTITIONING statement can be executed to convert a partitioned table that contains the 'UNIQUE CHECK IGNORE=1' attribute to a non-partitioned table.

• SQL statements are optimized for digest hash calculation when the statement outline feature is enabled and no outlines are available.

• The InnoDB on OSS feature is added.

• After the configurations of the statement outline and CCL features are updated on a multi-master cluster, the changes made on one primary node are automatically synchronized to other primary nodes.

• The error message that indicates spelling mistakes is optimized.

• The write performance of read/write nodes is optimized after semi-synchronous replication to standby nodes is enabled.

• The LIST DEFAULT HASH partitioned table type is added.

• The CREATE LIKE statement can be used to create LIST DEFAULT HASH partitioned tables.

• The without validation option is supported in the add partition syntax used to create LIST DEFAULT HASH partitioned tables and can be used to add list partitions.

• The SQL Trace feature can be used to track the execution information of slow queries.

• The lock granularity of Polar File System (PolarFS) is optimized to significantly improve performance in some high-load scenarios.

2023-06-12

Fixed issue

• The following issue is fixed: JSON files cannot be read by OSS.

• The following issue is fixed: Hint plans fail to be generated in ps mode.

• The following issue is fixed: Checkpoint LSNs are not pushed down as expected.

• The following issue is fixed: No data is found in the view_definition column of a system view.

• The following issue is fixed: An error may be reported when the ALTER TABLE ... IMPORT TABLESPACE statement is executed on read/write nodes and read-only nodes are queried.

• The following issue is fixed: Data that contains \n cannot be read after the DLM feature is used to archive the data.

• The following issue is fixed: Temporary files may remain in OSS after cold data is archived.

• The following issue is fixed: When OSS cold data in the CSV format is read, only the first file may be retrieved.

• The following issue is fixed: When slow query logs are concurrently recorded, the failure of the internal counter in the thread pool may cause a backlog of SQL requests.

• The following issue is fixed: Conflicts occur in auto increment columns when multiple INSERT commands are run on a multi-master cluster in parallel.

• The following issue is fixed: When the redo logs of a wide table are replayed on read-only nodes, replication latency occurs.

• The following issue is fixed: When the LIMIT OFFSET pushdown feature is used, invalid query results are returned for SQL statements that contain window functions.

• The following issue is fixed: When the CHECK TABLE statement is executed on a multi-master cluster, a lock needs to be held for the entire duration of the operation.

• The following issue is fixed: The source files stored in OSS cannot be deleted when the ALTER TABLE engine = csv storage oss, force; command is run.

• The following issue is fixed: Multi-master clusters crash when the caches of the clusters are dynamically adjusted.

• The following issue is fixed: The tablespaces of multi-master clusters cannot be exported.

• The following issue is fixed: When the engine_condition_pushdown parameter is set to true, tables that contain NULL fields are not pushed down correctly.

• The following issue is fixed: When the fast query cache feature is enabled, incorrect result sets are returned when the Time and Timestamp data types are used for comparison during a query.

• The following issue is fixed: After PolarTrans is dynamically enabled, database performance is degraded in high-concurrency update scenarios.

Category

Description

Release date

New feature and feature optimization

• Columns can be added in seconds by recreating IMCI data.

• The REGEXPR function is supported by IMCIs.

• The IS NULL and IS NOT NULL conditions are supported in the IMCI pruner feature.

• The cold data archiving feature is supported in Multi-master Cluster (Database/Table) Edition.

• When the minor version of a Multi-master Cluster (Database/Table) Edition cluster is updated, the minor version of all nodes in the cluster can be updated in parallel.

• Persistent statistics are supported in Multi-master Cluster (Database/Table) Edition.

• The optimizer trace generated when OSS pushdown is enabled for OSS foreign tables is optimized.

• Binary logs are not recorded when cold data is archived manually and automatically.

• The Innodb_scc_timeout_count parameter is added to record timeout events during global consistency (high-performance mode) queries.

• The Innodb_scc_timeout_degrade_count parameter is added to record the number of events when global consistency (high-performance mode) queries are converted to asynchronous queries in case of timeout.

• Global transaction identifiers (GTIDs) can be queried on read-only nodes.

• The following hint types are added for hint plans: BNL, BKA, and multi-range read (MRR).

• The qc_weak_read_time variable is added for the fast query cache to retrieve data from a cache without requiring the data to be the most up-to-date or real-time information.

• The dbms_outln.add_optimizer_outline('schema', 'hints', 'query') operation is supported for the statement outline feature to create optimizer hints.

• The auto plan cache feature is added.

• In SQL Sharing, the plan_extra, ref_by, and extra fields are displayed in the JSON format.

• Execution plans are displayed in the extended data of slow query logs.

• If the index hints specified for statement outline do not exist, only an alert is reported instead of an error.

• The by_expr_ignore_charset variable is added to SQL Sharing. The differences in the digests generated by the ORDER BY or GROUP BY statement in different character sets are eliminated after the by_expr_ignore_charset variable is set to ON. Setting by_expr_ignore_charset to ON results in different digest values for queries compared to when the parameter is set to OFF.

• Memory usage is optimized when tables in the recycle bin are scanned.

• The database engines of various versions are supported.

2023-04-14

Fixed issue

• The following issue is fixed: When an empty OSS foreign table is checked, the database node crashes.

• The following issue is fixed: An error is occasionally reported when the global read-only node of a Multi-master Cluster (Database/Table) Edition cluster is queried.

• The following issue is fixed: The sizes of metadata tablespaces for Multi-master Cluster (Database/Table) Edition clusters are not synchronized.

• The following issue is fixed: The custom table data in the system tablespace is rolled back for Multi-master Cluster (Database/Table) Edition clusters.

• The following issue is fixed: Undo logs cannot be recycled after the undo log recycling feature is enabled.

• The following issue is fixed: A read timeout error occurs and the database cannot be connected when physical replication is established on a read-only node.

• The following issue is fixed: Nodes crash after the ALTER TABLE DISCARD tablespace statement is executed on a Multi-master Cluster (Database/Table) Edition cluster, and then the ADD INDEX or DROP INDEX statement is executed on the cluster.

Category

Description

Release date

New feature and feature optimization

• The DECIMAL data type is supported for the IMCI feature.

• When an IMCI is created, data can be sorted based on specified columns to improve pruning efficiency.

• IMCIs improves data scanning for the ORC or PARQUET objects stored in OSS.

• Cost-based optimization can push the outer join conditions of a materialized derived table down to the inside of the derived table.

• The ability of the optimizer to estimate the number of scanned rows in associated columns in equal-cost inference scenarios is boosted.

• The join elimination feature of the optimizer can eliminate inner tables of left joins, including inner tables with multi-layer nested joins and semi-joins.

• The CAST(... AS INT) function is supported. This function is used in the same way as the CAST(... AS SIGNED INT) function.

• Mutual backup is supported for Multi-master Cluster Edition.

2023-02-21

Fixed issue

• The following issue is fixed: The database performance may be degraded in long-running transaction scenarios.

• The compatibility issue between global consistency (high-performance mode) and parallel queries is fixed.

• The following issue is fixed: When the transparent tablespace transfer feature is enabled on Multi-master Cluster Edition, DDL statements cannot be executed on other nodes.

• The following issue is fixed: The free size of a table is displayed as 4 PB in some cases.

• The following issue is fixed: When multiple primary nodes of a Multi-master Cluster Edition cluster are restarted in parallel, some nodes cannot be started.

• The following issue is fixed: The Multi-master Cluster Edition clusters crash when the primary node for writing full-text indexes and IMCIs is switched over.

• The following issue is fixed: The database cache cannot be accessed when DDL operations are copied on tables for which foreign keys are created.

• The timeout parameter related to the physical replication connection is modified to fix the issue of stuttering when the physical server fails.

• The following issue is fixed: The table lock cannot be restored during crash recovery of a Multi-master Cluster Edition cluster.

• The following issue is fixed: During the process of adding a node, accessing the information_schema.tables table can potentially cause the current node to crash.

• The following issue is fixed: Dirty pages are lost during crash recovery of a Multi-master Cluster Edition cluster.

• The following issue is fixed: After a database is deleted, the Multi-master Cluster Edition cluster crashes when the primary node for data writing is switched over.

• The following issue is fixed: When the thread pool is enabled, the TempTable engine may access invalid memory.

Category

Description

Release date

New feature and feature optimization

• PolarDB for MySQL Standard Edition provides the PolarDB HA automatic switchover feature that is developed based on the Reservation feature of Non-Volatile Memory Express (NVMe) disks.

• The dbms_ccl.add_ccl_digest_rule stored procedure is supported for the CCL feature to match CCL rules based on digest values.

• The variables that start with an at sign (@) can be used as input values in built-in stored procedures.

• The semicolon (;) at the end of an SQL statement of a query parameter in the following built-in stored procedures is ignored: dbms_ccl.add_ccl_digest_rule, dbms_outln.add_index_outline, and dbms_outln.add_optimizer_outline.

• All UNION clauses are supported for statement outline.

• The usernames and passwords in the mysql.servers table are no longer stored and displayed in plaintext.

  Note

  The existing username and password of a server are not encrypted after the upgrade. You must manually delete the existing username and password, and then recreate a username and password for the server.

• The LIMIT OFFSET pushdown feature can be used when an independent partition in a partitioned table is accessed.

• The parallel DDL feature is supported by X-Engine Edition clusters.

• The kernel autonomy capacity is supported for the thread pool feature.

• The thread pool feature is compatible with CCL. After the thread pool feature is enabled, if your connection is blocked by CCL, the connection is recorded in a thread pool, and the thread pool creates additional threads to execute SQL statements on other connections.

• The OSS_REPLACE_EXISTED_FILE() hint is added to directly overwrite the existing file that has the same name as a specified file when the SELECT INTO OSSOUTFILE statement is executed.

• Single-node parallel execution can be used in OSS.

• The INNODB_GLOBAL_DDL_LOCK_INFO table is added to the information_schema database to display the global locks of a multi-master cluster.

• The AWR SQL Detail feature is supported to audit and manage DDL, LOCK DB, and LOCK TABLE statements.

• The ANALYZE TABLE command is run to collect statistics on IMCIs to help users make query decisions.

• The database/table restoration feature is supported for tables that contain IMCIs.

• The cast_as_real function is supported for the IMCI feature.

• The creation status and progress of IMCIs can be recorded in the INFORATION_SCHEMA.IMCI_ASYNC_DDL_STATS table.

• The performance of associated queries where the associated fields in IMCIs are strings are optimized.

2022-12-07

Fixed issue

• The following issue is fixed: IF NOT EXISTS cannot be included in the syntax that is used to create servers.

• The following issue is fixed: Exceptions may occur when DDL statements are executed on the primary nodes of a multi-master cluster.

• The following issue is fixed: Deadlocks may occur in a multi-master cluster when the tables of the information_schema database in the cluster are queried.

• The following issue is fixed: The replace performance of a multi-master cluster is degraded.

• The following issue is fixed: Global locks are required when implicit temporary tables are created for a multi-master cluster.

• The following issue is fixed: Multi-master clusters are in the pending state when the clusters recover from a crash.

• The following issue is fixed: The number of logic reads is incorrectly collected by the SQL Trace feature.

• The following issue is fixed: The output result is incorrect when IMCIs contain the ENUM and SET data types.

• The following issue is fixed: The SQL statements that contain ROLL UP and GROUP BY cannot be routed to column store nodes for execution.

Category

Description

Release date

New feature and feature optimization

• The UNION ALL operation is optimized for IMCI.

• Built-in stored procedures are supported for the IMCI feature to facilitate IMCI creation. For more information about the built-in stored procedures, see Overview.

• The imci_default_pack_shift global parameter is supported for the IMCI feature to specify the default pack size of an IMCI.

• Partition pruning can be used for the IMCI feature to optimize the query performance of partitioned tables.

• Data of the BIT, JSON, and GEOMETRY types is supported for IMCI.

• The EXCHANGE PARTITION operation is supported for IMCI.

• The performance of global consistency (high-performance mode) is improved by using the MTT technique.

• The adaptive elimination of data pages in the cache center is supported in Multi-master Cluster Edition.

• The recursion depth of a JSON file can be 150.

• The mm_flush_autoinc_when_unlock parameter is introduced in Multi-master Cluster Edition to specify whether to refresh auto-increment columns when the global locks on databases or tables are released. The default value of this parameter is false.

• If the Schema_name parameter is left empty in an SQL statement, and the digest value of the statement matches the digest value specified by a statement outline, the statement outline takes effect.

• The fast query cache feature can access the cache of the metadata table named information_schema.tables/columns in the database.

• The SQL Trace feature is added to track the execution process of SQL statements. For more information about the feature, see Usage.

• The SELECT INTO OSSOUTFILE statement is supported to export local tables on PolarFS to OSS.

2022-10-29

Fixed issue

• The following issue is fixed: Data is incompletely read when RDMA is used.

• The following issue is fixed: Table clean threads are abnormal when primary nodes are shut down.

• The following issue is fixed: Deadlocks occur when Offline Master commands are run and WAR operations are performed.

• The following issue is fixed: Primary nodes cannot be shut down when the nodes are disconnected from the cache center.

• The following issue is fixed: Nodes crash when data is inserted into tables with implicit primary keys by using the Event Scheduler.

• The following issue is fixed: Binary logs are generated when SQL statements are executed during hot upgrade.

Category

Description

Release date

Fixed issue

The following issue is fixed: Read-only nodes crash when the hash values of file names conflict.

2022-09-17

Category

Description

Release date

New feature and feature optimization

• Read-only nodes support the failover with hot standby feature.

• The ccl_mode parameter is supported for the CCL feature. If a CCL threshold is specified, when the access traffic of your database reaches the specified threshold, other requests sent to access the current cluster are denied.

• The oss_sts_token parameter is added to the statement for creating an OSS server to configure temporary credentials for OSS.

• The adaptive digest length is supported, which indicates that the memory space calculated by the statement digest function is adaptive to the length of query statements.

• Cluster endpoints can be used to query tables related to IMCI in INFORMATION_SCHAME.

• The extract, transform, and load (ETL) mode is added for the IMCI feature. You can use the data queried from read-only column store nodes to create tables and write the data to the tables.

• The performance of TopK queries that contain the order by a limit n,m clause is optimized for IMCI.

• The Last_stmt_use_imci state is added for the IMCI feature to check whether IMCIs are used in query statements.

• The polar_enable_implicit_imci_with_create_table global variable is supported for the IMCI feature to automatically create IMCIs when the CREATE TABLE statement is executed.

• Cursors are supported for the IMCI feature.

• IMCIs are supported for explicit transaction queries.

• IMCIs can be created on partitioned tables.

2022-09-08

Fixed issue

• The following issue is fixed: The cached content is inconsistent with the original object due to transaction rollback when the copy algorithm is used in DDL statements.

• The following issue is fixed: Slow logs are not recorded when a database is deleted.

• The following issue is fixed: Errors are returned when both LIMIT OFFSET pushdown and MRR optimization are enabled.

Category

Description

Release date

New feature and feature optimization

Thread pool optimization: When database transactions wait for metadata locks or row locks, additional threads can be generated in the thread pool to process new requests.

2022-08-12

Category

Description

Release date

New feature and feature optimization

• The IMCI feature is supported. The IMCI feature and the parallel vectorized execution engine for column store data can greatly improve the execution efficiency of complex queries. For more information, see Overview.

• The FEDERATED storage engine is supported. You can create and use FEDERATED tables.

• In-memory temporary tables can be customized on read-only nodes. Only the MEMORY and MyISAM engines are supported.

• Optimized DDL sorting: Two-way merging is changed to multiple-way merging to reduce I/O operations and improve the performance of index creation by using DDL statements.

• The read-ahead mechanism for DDL statements is supported, which improves the efficiency of index creation.

• The oss_prefix parameter is added to the statement for creating an OSS server to configure the OSS object path.

2022-07-20

Category

Description

Release date

Fixed issue

The following issue is fixed: High latency occurs when physical replication is performed during a time period of low workloads in the cluster.

2022-05-25

The following issue is fixed: When a large number of dirty pages exist in the buffer pool and page buffer pool flushing is slow, the least used page cannot be evicted.

Memory allocation during data insertion is optimized.

The following issue is fixed: If the innodb_show_verbose_deadlock parameter is set to the default value ON, the full deadlock information is queried after the show engine innodb status command is run.

The following issue is fixed: New read-only nodes in a cluster crash after the UNDO TRUNCATE statement is executed on an existing read-only node.

The following issue is fixed: When a skip scan is performed and some data in the database is marked as deleted, the results of the skip scan may be incorrect.

Category

Description

Release date

Fixed issue

The issue of incorrect results for using index for group by is fixed.

2022-05-05

The following issue is fixed: When a skip scan is performed and some data in the database is marked as deleted, the results of the skip scan may be incorrect.

The following issue is fixed: When a skip scan is performed on an index that is prefixed with multiple sets of data, the results of the skip scan may be incorrect.

Category

Description

Release date

Fixed issue

The issue of incorrect results for using index for group by is fixed.

2022-04-12

Category

Description

Release date

Fixed issue

• The following issue is fixed: threadpool_high_priority_users is not properly parsed when the database starts.

• The following issue is fixed: Dynamic modification of threadpool_high_priority_users may affect other parameters.

• The following issue is fixed: Redo logs are not replayed on the read-only nodes of X-Engine.

2022-03-11

Category

Description

Release date

New feature and feature optimization

• Strict consistency cluster is supported on read-only nodes.

• PrepareStatement is supported in the statement outline feature.

• The hot standby node feature is provided to further optimize HA efficiency.

2022-01-20

Fixed issue

• The following issue is fixed: The fast query cache feature blocks the synchronization of redo logs when the read-only node obtains the metadata lock (MDL).

• The SELECT FOR UPDATE/SHARE WAIT N statement is supported. N specifies the number of seconds during which a row remains locked. The timeout applies only to locking a single row. To lock multiple rows for a query, the system detects the timeout only based on a single row instead of multiple rows. If the lockout time exceeds the specified N value, the system returns the Lock wait timeout exceeded; try restarting transaction error message.

Category

Description

Release date

Fixed issue

• The estimation for the degree of parallelism (DOP) of parallel execution plans is more accurate when the indexes of partitioned tables are the same.

• The following issue is fixed: The optimizer takes a long time to estimate the costs of statistics in some scenarios.

• The following issue is fixed: The optimizer fails to choose a better path to the range of an index for some GROUP BY statements.

• The following issue is fixed: After a read-only node is promoted to a primary node by using a HA architecture, redo log files fail to be created.

2021-12-22

Category

Description

Release date

Fixed issue

• The following issue is fixed: When a large value is specified for group_concat_max_len, an arithmetic overflow occurs, which causes an incorrect result of the GROUP_CONCAT function.

• The following issue is fixed: Data errors occur on read-only nodes in a zone after the nodes recover from a failure.

• The following issue is fixed: Data exceptions occur after a read-only node in a zone is promoted to the primary node.

• The following issue is fixed: After the filter conditions are specified for multi-column indexes, the optimizer fails to filter the columns and can only filter single-column indexes.

2021-11-26

Category

Description

Release date

New feature and feature optimization

• New concurrency control rules are added for DDL operations.

• The restrict_on_limit_without_order parameter can be used to specify whether parallel queries are allowed for statements that contain the LIMIT clause and do not contain the ORDER BY clause.

2021-09-18

Fixed issue

• The following issue is fixed: When an index is selected to perform GROUP BY operations on columns in scenarios where implicit GROUP BY sorting is used in parallel queries, the results of concurrent execution are not sorted.

• The following issue is fixed: When the thread pool is used, the transaction ID field in audit logs is always 0.

Category

Description

Release date

New feature and feature optimization

Transaction resumable upload is supported for the database kernel. Uncommitted transactions can be read by read-only nodes for the database kernel.

2021-08-14

Fixed issue

master_key_id_mutex is optimized to ensure that DDL operations can be performed in parallel.

Category

Description

Release date

New feature and feature optimization

• The polar_replica_work_on_nonblock_mdl_mode parameter is added. If this parameter is set to on, the uncommitted transactions that use the READ UNCOMMITTED or READ COMMITTED isolation level on read-only nodes no longer block DDL operations on the primary node. In addition, atomicity, consistency, isolation, and durability (ACID) is no longer ensured for table creation on read-only nodes.

• The mechanism for maintaining statistical information is optimized for scenarios in which a large number of tables need to be used, such as scenarios in which software-as-a-service (SaaS) services are required. This way, the performance of querying table schema information by executing the DESC TABLE statement in high concurrency scenarios is significantly improved. The read and write performance in high concurrency scenarios is also significantly improved.

2021-07-23

Fixed issue

• The following issue is fixed: A read-only node crashes due to m_recv_bits.is_set(fold) when physical replication is being performed on the node in high concurrency scenarios.

• The following issue is fixed: A read-only node crashes when the node replays logs in high concurrency scenarios.

Category

Description

Release date

Fixed issue

When requests are in the Waiting for acl cache lock state, the information about the request threads and the first thread that holds the lock can be printed to the master error logs. This facilitates issue diagnostics.

2021-06-24

Category

Description

Release date

New features and optimized features

• The TDE feature can be used to automatically encrypt data in tables that are newly created in PolarDB for MySQL clusters.

• MySQL data tables support the utf8mb4_0900_bin character set.

2021-05-25

Fixed issue

• The following issue is fixed: Databases crash due to an excessive number of log records that are generated when the Instant ADD COLUMN feature is used and then an update operation is rolled back.

• The following issue is fixed: Errors occur when DDL statements are executed to forcibly change the storage engine of the mysql.slow_log and mysql.general_log tables to InnoDB.

• The following issue is fixed: Inaccurate metadata about the REGEXP function causes invalid result sets.

• The following issue is fixed: The database crashes when an update operation is rolled back on a virtual column.

• The policy that specifies whether the checkpoint process is immediately triggered after read-only nodes establish a replication relationship with the primary node can be adjusted. If the difference between the LSNs of two consecutive log records is less than the specified threshold, a policy can be configured to prevent the checkpoint process from being triggered.

Category

Description

Release date

New feature and feature optimization

• The mechanism for maintaining internal index information is optimized for scenarios in which a large number of tables need to be used, such as scenarios in which SaaS services are required. This way, read-only nodes can be started within a shorter time period.

• Synchronization links in global database networks (GDNs) are optimized. This reduces the amount of CPU resources consumed by synchronization threads. This also improves the speed of synchronizing redo logs between clusters that have less than 8 CPU cores.

• The AutoDop policy specified by the parameter that controls the DOP is optimized to prevent the impact caused by the usage of PARALLEL HINT and force_parallel_mode.

2021-04-23

Fixed issue

• The following issue is fixed: When a range query covers a large scope, invalid indexes are used due to the inaccurate statistics returned by records_in_range.

• The following issue is fixed: Read-only nodes crash due to excessively small buffer pools when a point-in-time restore is used to restore full data.

• The following issue is fixed: The memory usage of metadata caches in X-Engine is excessively high.

• Two ACL DDL issues in MySQL are fixed to prevent cluster unavailability caused by deadlocks that occur due to ACL DDL operations.

• The following issue is fixed: Invalid result sets are returned when aggregate functions are used in parallel queries that include the SQL_BUFFER_RESULT keyword and exclude GROUP BY.

Category

Description

Release date

Fixed issue

The following issue is fixed: The encryption information in tables that are encrypted by TDE may be lost during a failover. This results in decryption errors.

2021-04-19

Category

Description

Release date

Fixed issue

The following issue is fixed: For parallel queries, a longer response time is required because constant filter conditions specified in Block Nested-Loop (BNL) join algorithms are not pushed down to each queried table.

2021-04-08

Category

Description

Release date

New feature and feature optimization

• The polar_use_statement_mdl_on_replica parameter is available for read-only nodes. When this parameter is set to on, this setting prevents transactions that use the RC isolation level on read-only nodes from blocking DDL operations on the primary node. When a read transaction on a read-only node is concurrent with a DDL operation on the primary node, the statements in the read transaction read different data sets from tables. For example, if two statements are included in a read transaction on a read-only node and an ADD COLUMN operation is performed on the primary node during the execution interval of the two statements, the second statement on the read-only node reads data from more columns than the first statement.

• Unnecessary equality conditions for indexes in queries are removed. This pushes down the LIMIT and OFFSET clauses when the ORDER BY LIMIT clause is used.

• The dbms_stats.gather_stats(timeout, workers) command is added. You can use scheduled events or run the command to update outdated histograms.

• The log_version, log_id, origin_start_time, and rds_ext fields can be queried from the mysql.slow_log table.

2021-03-30

Fixed issue

• The following issue is fixed: Queries cannot be canceled when CHECK TABLE, COUNT(*), or DDL statements are executed on X-Engine tables.

• KICKOUT is used as a non-reserved keyword.

• The following issue is fixed: In some scenarios, parallel queries are not performed as expected when the evaluated number of rows to be scanned in the execution plan is less than the actual number of scanned rows.

Category

Description

Release date

Fixed issue

• The following issue is fixed: If the session_track_temporary_tables system variable is set to on, the cluster becomes unavailable when temporary tables are created or deleted within a stored procedure.

• A patch for MySQL 8.0.14 is provided to fix the following issue: CREATE USER statements cannot be executed as expected because the MDLs of system tables in the corresponding MySQL database cannot be obtained.

2021-03-12

Category

Description

Release date

New feature and feature optimization

The write performance of X-Engine in PolarDB is optimized when data is continuously imported to multiple tables.

2021-03-02

Fixed issue

The following issue is fixed: In parallel queries, the MDL state of the leader thread may be different from that of the worker thread, because the leader thread does not own mutual exclusion locks.

Category

Description

Release date

New feature and feature optimization

• The auto_dop_low_degree_cost parameter is added to specify the DOP policy. This parameter serves as a threshold that determines the cost at which the query optimizer selects a lower DOP. For more information, see Configure DOP policies.

• The restore_table command is supported. You can run the command to restore tables from the recycle bin. For more information, see Table recycle bin.

• Binary logs can be obtained from read-only nodes. For more information, see Remotely obtain and parse binary logs of a PolarDB for MySQL cluster.

• Key information such as in_memory can be displayed in opt trace. This can be beneficial for troubleshooting and pinpointing the causes of issues in query execution plans.

2021-02-20

Fixed issue

• The Port Percona patch is provided to reduce the time that is required to detect conflicts of ACL cache MDLs in high concurrency scenarios.

• The preferred_ordering_index parameter is added to the optimizer_switch system variable to fix the following issue: The optimal plan is not used due to ordered indexes in some scenarios. For example, this issue occurs when LIMIT clauses are used in ORDER BY or GROUP BY queries.

• The following issue is fixed: The SHOW PROCESSLIST statement in some scenarios returns an invalid result.

• The following issue is fixed: For PolarDB for MySQL 8.0.1.1.10 or earlier, results cannot be quickly returned if system tables are queried after a minor version update. This is because the definition of the information_schame.KEY_COLUMN_USAGE table is not updated.

Category

Description

Release date

New feature and feature optimization

• The parallel_degree_policy parameter can be used to specify the DOP policy for parallel queries. For more information, see Configure DOP policies.

• The SET GLOBAL statement can be executed to specify the max_digest_length parameter. This parameter is used to limit the length of recognizable statements.

  Note

  After the value of max_digest_length is changed, the client must reconnect to the cluster. Otherwise, the new value does not take effect.

• The default value of the innodb_adaptive_hash_index parameter is changed from ON to OFF.

2021-01-29

Fixed issue

• The following issue is fixed: The permissions on the primary node are different from the permissions on read-only nodes.

• The following issue is fixed: Read-only nodes cannot connect to the primary node after a failover.

• The following issue is fixed: The processing logic of the SPM is incorrect when a specific execution plan becomes invalid.

Category

Description

Release date

New feature and feature optimization

• The implicit GROUP BY sorting feature is supported. This feature works in the same way as in PolarDB for MySQL 5.7.

• Parallel queries can be disabled if a BLOB field exists.

• The cache information of statement concurrency control can be automatically updated on read-only nodes.

• The hot row optimization feature is supported. For more information, see Hot row optimization.

• DDL physical replication optimization is supported. For more information, see DDL physical replication optimization.

• Async metadata lock replication is supported. For more information, see Async metadata lock replication.

• Quick reverse traversal is supported when computations are pushed down to InnoDB.

• The file system is optimized to speed up the process of opening multiple tables at the same time.

• In multi-table scenarios, the time consumed by a failover is reduced and the recovery of the new primary node is accelerated.

2021-01-12

Fixed issue

• The following issue is fixed: The system tables are lost when a read-only node is promoted to the primary node.

• The following issue is fixed: The estimated number of scanned rows is excessively large when range queries are used after the parallel query feature is enabled.

• The following issue is fixed: The results of aggregate queries are of the integral data type if the field data type is BIT.

• The following issue is fixed: SELECT DISTINCT queries return invalid results after enumerated fields are used.

• The following issue is fixed: An exception occurs in the results of parallel queries that use the EXISTS condition.

• The following issue is fixed: Read-only nodes fail to restart in some scenarios.

• The following issue is fixed: An exception occurs in table information in the data dictionary when a read-only node executes DDL statements. This is because the table for which DDL statements are being executed is re-opened when a foreign key association table is opened.

• The following issue is fixed: When full-text indexes are used to query data, failures occur due to the improper setting of the node restart flag after a failover.

• The following issue is fixed: The log application threads on read-only nodes are blocked due to an MDL.

• The following issue is fixed: The new primary node becomes unavailable after a failover because the released memory is reused.

• The following issue is fixed: All nodes become unavailable due to polar.info data errors.

• The following issue is fixed: Auto-increment column exceptions occur in partitioned tables.

• The following issue is fixed: A data error occurs when redo logs on the primary node are overwritten.

• The following issue is fixed: The primary node becomes unavailable when it is waiting for an MDL.

• The issues related to the TDE feature are fixed.

• The following issue is fixed: Clusters become unavailable when the LOCK TABLE statement is executed and the recycle bin feature is enabled.

• The following issue is fixed: A deadlock occurs when DDL statements are executed on the primary node.

• The following issue is fixed: The thread pool and the connection control cannot take effect at the same time.

Category

Description

Release date

New feature and feature optimization

SPM and PLAN are removed from the keyword list. This ensures that tables whose names contain SPM or PLAN can be managed without conflicts with reserved keywords.

2020-12-18

Category

Description

Release date

New feature and feature optimization

• The multi-plan mode is supported in the execution plan manager.

• The rds_ap_threshold system variable is added to block the requests that cause the system to scan an excessive number of records. These requests are evaluated by the optimizer.

• The efficiency of writing dirty pages to disks on the primary node is improved.

• A multi-shard writing mechanism for redo logs is supported.

2020-12-09

Fixed issue

• The following issue is fixed: A null pointer of the MDL key occurs in parallel queries.

• The following issue is fixed: Queries fail when a parallel thread cache is created.

• The following issue is fixed: An error is returned when MRR is used in parallel queries.

Category

Description

Release date

New feature and feature optimization

• The efficiency of scanning driven tables in parallel is improved in scenarios such as JOIN queries.

• When the binary logging feature is disabled, the remaining binary logs can be cleared.

• The interrupted physical replication of read-only nodes can be automatically detected and then resumed. This way, physical replication can be resumed at the earliest opportunity when it is interrupted.

• The efficiency of failovers between the primary node and read-only nodes is improved.

• Quick startup is supported for clusters that contain a large number of tables. This way, data files can be rapidly scanned.

2020-11-16

Fixed issue

• The following issue is fixed: A cluster may crash when the type of trx->wait_lock is obtained.

• The following issue is fixed: The number of AIO threads reaches the upper limit when a multi-queue simulated AIO is enabled.

• The following issue is fixed: A query cannot be canceled if a query initialization failure occurs when indexes are queried.

• The following issue is fixed: The Next Page of the current cursor points to a page that does not exist when a read-only node is performing a split-merge operation (SMO).

• The following issue is fixed: Read-only nodes read logs that are overwritten by the primary node.

• The following issue is fixed: The system fails to clear redo logs because the timestamp intervals of redo logs are excessively large.

• The following issue is fixed: The table cache information in the related cache is not cleared after MDLs are released.

Category

Description

Release date

New feature and feature optimization

• The compatibility between the SPM and parallel queries is improved.

• The efficiency of merging and sorting parallel queries is improved.

• DELETE operations can be pushed down to InnoDB.

• The Commit Timestamp (CTS) feature of PolarDB is supported.

2020-09-21

Fixed issue

• The following issue is fixed: The description of pq_optimize_switch is incorrect.

• The following issue is fixed: Subqueries cannot be stably executed.

Category

Description

Release date

New feature and feature optimization

• A feature is added to decide whether the primary node must immediately apply the checkpoint policy when it establishes a replication relationship with read-only nodes.

• Range queries are supported and can be pushed down.

• The pfs_remount feature is supported for Polar File System (PFS). This feature fixes the issue that files in PFS fail to be mounted when the files are open.

• The performance bottlenecks caused by forcible pauses of parse threads on read-only nodes are removed. The efficiency of data synchronization between the primary node and read-only nodes during physical replication is improved.

• The performance of Early Lock Release (ELR) is improved in multi-connection scenarios. The cluster performance in multi-connection scenarios is improved by nine times.

2020-08-19

Fixed issue

• The following issue is fixed: Read-only nodes become unavailable after these nodes fail to connect to the primary node.

• The following issue is fixed: After a full-text index is used and DDL statements are executed, a read-only node becomes unavailable after a failover.

• The following issue is fixed: Binary logs cannot be purged after the UNDO TRUNCATE command is run.

• The following issue is fixed: Statistics are inconsistent between the primary node and read-only nodes.

Category

Description

Release date

New feature and feature optimization

• DDL statements can be executed in parallel. This improves the efficiency of executing DDL statements.

• The length of each queue for the simulated asynchronous I/O (AIO) can be dynamically adjusted.

• Data in a full-text search (FTS) index cache can be synchronized between the primary node and read-only nodes. This ensures cache coherence.

• The subqueries that include aggregate functions are supported in WHERE clauses. In addition, if the subqueries support index-based scans, these subqueries can be executed in parallel.

• The lock mode can be checked for temporary tables and standard tables.

2020-07-04

Fixed issue

• The following issue is fixed: Clusters become unavailable because some DDL statements are being replicated when the primary node is downgraded to a read-only node.

• The following issue is fixed: The performance is compromised because thread pools are enabled.

• The following issue is fixed: Deadlocks occur because binary logs are purged.

• Several issues related to a memory leak are fixed.

• Several issues that occur in high availability scenarios are fixed.

Category

Description

Release date

New feature and feature optimization

• Security is enhanced. For example, password management is improved.

• The performance of parallel queries is improved in the following scenarios:

  • The GROUP BY, UNION, and SELECT COUNT(*) FROM <Table> queries are performed.

  • Shared InnoDB temporary tables are used for execution plans in parallel subqueries.

  • Views, derived tables, and temporary tables are used in execution plans.

  • Temporary tables can be used for parallel queries. When temporary tables are used for parallel queries, take note of the following limits:

    • If conditions are not specified in the SELECT COUNT(*) statement when data in temporary tables is queried, parallel queries are not supported.

    • Parallel queries are not supported for temporary tables that use the Memory storage engine.

• A new version of the audit log format is supported. In this format, information about virtual IP addresses (VIPs) is added.

• The rate of idle space to total space on index pages can be managed. This reduces latch contention and the probability of performing structure modification operations (SMOs), and improves write performance.

• Simulated AIO among multiple queues is supported. This improves dirty page flushing and write performance.

• Buffer pool pages can be excluded from core files. This reduces the sizes of the core files and minimizes the impact on online services.

2020-05-29

Fixed issue

• The following issue is fixed: If the upper limit of memory in TempTable is reached, the TempTable storage engine reports an out of memory (OOM) error. In this case, the data is not moved from memory to disks.

• The following issue is fixed: Clusters become unavailable when the sort buffer size is set to an excessively small value and ORDER BY is used in InnoDB FTS.

• The following issue is fixed: If a temporary table contains duplicate column names, the requested fields fail to be returned.

• The following issue is fixed: In parallel queries, when the MAX or MIN function is used together with GROUP BY and loose index scans are used, the statements are still being executed after the queries are killed.

• Several issues related to failovers are fixed.

• Several issues related to parallel queries are fixed.

• The following issue is fixed: When the SHOW BINARY LOGS statement is executed, transactions may not be committed.

Category

Description

Release date

New feature and feature optimization

• The sorting method is optimized. Column values are sorted based on string prefixes, and then based on entire strings. Specifically, if the string prefixes of the column values are the same, the column values are sorted based on the entire strings. When columns of the CHAR and VARCHAR data types are sorted, the maximum length of different strings in the prefix can be specified to accelerate the comparison process. This reduces the sorting time.

• Parallel queries are supported in the following scenarios:

  • Parallel queries are supported for cost models of range scans.

  • Parallel queries are supported for temporary tables.

  • Parallel queries can be performed based on the semi-join materialization strategies: lookup and scan.

• Three types of session state trackers are added. The PolarDB intelligent proxy can use these session state trackers to enable the persistent connection feature. After the trackers are enabled, the changes of user variables in sessions, CREATE and DELETE operations on temporary tables, and PREPARE and DEALLOCATION operations in SQL statements can be tracked.

• The performance of dropping an adaptive hash index (AHI) is optimized when DDL statements are executed. This reduces the impact of DDL statements on cluster performance.

• The recycle bin feature is added to prevent data loss that is caused by unintended deletion operations.

• The performance of truncating temporary tablespaces in large buffer pools is improved. This reduces the impact of the operations related to temporary tables on cluster performance.

2020-04-09

Fixed issue

• The following issue is fixed: If aggregate functions are nested in the IF function, an error occurs when the ROLLUP function is executed.

• The issues that occur when the values of the BLOB data type are sorted are fixed.

• The issue that occurs when the SQL PREPARE statement includes aggregate functions in parallel queries is fixed.

• Several issues related to parallel queries are fixed.

• The following issue is fixed: Excessive redo logs may be cleared.

• Issues related to the redo logs of read-only nodes are fixed.

Category

Description

Release date

New feature and feature optimization

• Parallel queries are supported when subqueries contain the ROLLUP function.

• The statement concurrency control feature is supported.

• The POLARDB_INDEX hint is added.

• The synchronization latency between the primary node and read-only nodes is reduced.

• The thread pool feature is supported. For more information, see Thread pool.

• The keyring_rds plug-in for TDE is supported.

• GDNs are supported. For more information, see Create and release a GDN.

• The lock-free transaction system is optimized and read and write performance is improved.

2020-03-28

Fixed issue

• Several issues related to parallel queries are fixed.

• The following issue is fixed: A value of 0 may be returned for the statistics in online DDL operations.

• The file system in user mode is optimized to accelerate cluster startup.

• The following issue is fixed: Clusters may become unavailable when the innodb_flush_method parameter is set to all_o_direct.

• The following issue is fixed: Clusters may become unavailable if a lock is released after a transaction is committed.

• The following issue is fixed: User requests may be blocked when slow query logs are truncated.

• The following issue is fixed: Clusters may become unavailable due to the compressed pages on read-only nodes.

• The following issue is fixed: Replication connections may be closed in thread pools by mistake.

Category

Description

Release date

New feature and feature optimization

• The parallel query feature is improved, and the parallel computing of the ROLLUP function that is used for enterprise-level analysis is supported.

• The capabilities of the optimizer cost model are improved. Filters can be specified to decrease selectivity. The cost model is supported for parallel queries. The optimizer checks whether a parallel or serial execution plan is used for the executed SQL statements based on selectivity.

• Parallel threads can be allocated in first in, first out (FIFO) mode and managed in a centralized manner. This ensures that system resources are not exhausted by a large number of parallel queries.

2020-02-05

Fixed issue

• Issues related to memory in parallel queries are fixed.

• Issues related to instability in parallel queries are fixed.

Category

Description

Release date

Fixed issue

• The following issue is fixed: The binary log index file is not closed after the primary node is downgraded to a read-only node.

• The following issue is fixed: Clusters become unavailable when read-only nodes access undo log pages that are purged.

• The following issue is fixed: Background threads access tablespace pages that do not exist when a read-only node is promoted to the primary node during a failover.

• The following issue is fixed: After clusters are shut down, log threads exit but redo logs are still being written. As a result, the clusters become unavailable.

2020-01-01

Category

Description

Release date

New feature and feature optimization

• The optimizer trace feature can be used to track information about parallel queries. For example, the information facilitates the analysis of the reason why a parallel query is used or not used.

• Hints are added for parallel queries. You can add hints to SQL statements to explicitly enable parallelism and specify the DOP.

• Parallel scans are supported if the INSERT INTO SELECT statement is executed when the isolation level is READ COMMITTED. The INSERT INTO SELECT statement can be used to query data and insert the result into another table.

2019-12-03

Fixed issue

• Several issues related to parallel queries are fixed.

• The following issue is fixed: After a read-only node is promoted to the primary node during a failover, the new primary node becomes unavailable.

• The following issue is fixed: Errors are reported due to the execution of some DDL statements when a failover occurs.

• The following issue is fixed: The "too many connection error" error message is reported when the upper limit of connections that are supported by locks is reached.