ApsaraDB for Redis:ApsaraDB for Redis proxy nodes

• Timed-out read requests can be automatically forwarded to other read replicas.
• TLS 1.3 is supported.
• The TFT.ANALYZER and TFT.EXPLAINCOST commands of TairSearch are supported.
• Command outputs are changed from the encapsulated results on proxy nodes to the execution results from databases.
• The computing logic of channels upon incremental subscription is optimized to reduce CPU utilization.

• The issue concerning a superfluous space at the end of one-line audit logs is fixed.
• The issue concerning memory leaks that may occur when the client uses an incorrect protocol is fixed.

• Commands of Redis 6.2 and 7.0 are supported.
• The TFT.ANALYZER command of TairSearch is supported.

• TLS 1.3 is supported.
• The TFT.ANALYZER and TFT.EXPLAINCOST commands of TairSearch are supported.

• The issue concerning a superfluous space at the end of one-line audit logs is fixed.

• TairVector is supported.

• The issue that may cause filter aggregation errors in TairSearch is fixed.
• The following issue is fixed: After the ptod_enabled parameter is set to 1, the IP address of the client in audit logs becomes incorrect.

• The OS field is added to the output of the INFO command.
• The CLIENT KILL user command is supported.

• The issue that may cause incomplete returned data of the MOVED command is fixed. This prevents the failure of resolving the client protocol.

• The BF.INFO command is supported.
• The EXHSCANUNORDER command of TairHash is supported.
• The maximum size of an audit log is changed from 4 KB to 2 KB.

• The issue concerning incorrect IP addresses in audit logs of cloud disk-based instances in proxy mode is fixed.

• Stability is enhanced.

• Specific TairSearch commands are supported.
• The authentication in the AUTH user:password format is supported.

• The issue concerning nested empty arrays caused by the REdis Serialization Protocol (RESP) V3 is fixed.

• The RESP V3 can be resolved and forwarded, and the communication mode between proxy nodes and Tair instances can be changed by using the resp_version parameter.

• The stability is improved, and specific crashes are fixed.

• Specific TairSearch commands are supported.
• New commands of TairRoaring V2.2 are supported.

• Specific TairZset commands are supported.
• Specific TairRoaring commands are supported.
• The RC4 encryption algorithm is disabled for SSL certificates.

• The following issue is fixed: After the ptod_enabled parameter is enabled, an exception occurs when the SDIFFSTORE, SINTERSTORE, SUNIONSTORE, ZINTERSTORE, or ZUNIONSTORE command is run.
• The following issue is fixed: The CROSSSLOT error occurs when the SMOVE command is run.

• The following issue is fixed: \n in the command output is truncated when the DBSIZE or KEYS command is run on instances that have abnormal nodes.

• The following issue is fixed: The SCRIPT DEBUG command cannot be used.
• The following issue is fixed: The output score of the ZINTERSTORE or ZUNIONSTORE command has only six decimal places.

• Stability is enhanced.

• Stability is enhanced.

• The RANDOMKEY command is optimized to retrieve a different random node each time. This way, if the RANDOMKEY command is run multiple times, it does not retrieve the same node repeatedly.

• The following issue is fixed: The aggregations that derive from info commandstats on DRAM-based instances encounter errors.

• Stability is enhanced.

• The following issue is fixed: After an SSL connection is established for an instance, the first sent request may not receive a response.

• Stability is enhanced.

• Stability is enhanced.

• The following issue is fixed: In the Elastic Compute Service (ECS) architecture, the ZINTERSTORE and ZUNIONSTORE commands may fail to return responses when the split_multi_key_cmd_as_slot parameter is enabled.

• The following issue is fixed: After the proxy query cache feature is enabled, hot upgrades may fail.

• The stability is enhanced.

• The stability is enhanced.

• The following issue is fixed: An error message is returned when an ApsaraDB for Memcache instance processes only read or write requests.

• The CVE-2021-3711 and CVE-2021-3712 vulnerabilities are fixed.

• The stability is enhanced.

• Memory leaks that occur when the statistics feature is enabled are fixed.

• The stability is enhanced.

• The Memcached gateway mode is supported. This allows you to use the Memcached protocol to forward requests.

• The CLIENT LIST or CLIENT KILL command can be used to query or manage client connections by process.

• The issue that TairZset commands do not support uppercase letters is fixed. For more information, see TairZset.

• Internal management in cross-zone disaster recovery scenarios is optimized.

• More internal commands of ApsaraDB for Redis are available.

• The TairZset data structure is added. It allows you to sort score data of the DOUBLE type with respect to different dimensions. This data structure improves the data processing efficiency and is easy to use on the client side because you do not need to encode, decode, or encapsulate data. For more information, see TairZset.

• The proxy query cache feature is added. After you enable this feature, proxy nodes cache the request and response data of hotkeys. If a proxy node receives a duplicate request within the validity period of cached data, the proxy node directly returns a response to the client without the need to interact with backend data shards. This feature reduces access skew that occurs when a large number of read requests for hotkeys are sent. For more information, see Use proxy query cache to address issues caused by hotkeys.

• The following issue is fixed: Infinite loops occur when the commands used to manage multiple keys are run in special scenarios.

• The following issue that occurs in 6.5.5 is fixed: Requests are out of order when multiple databases are requested. This minor version is a special version released based on 6.5.5.

• The following issue is fixed: Requests are out of order when multiple databases are requested.

2021-04-09

• The maximum number of data shards supported by the SCAN command increases from 256 to 1,024.
• If the slots of the channels to which clients subscribe are migrated, proxy nodes close the client connections and reconnect to the clients. This ensures data consistency.

• Requests are resent to the address specified by the MOVED command when this command is run.
• Proxy nodes do not route the commands without specified keys to the data shards that have empty slots.

• The following issue is fixed: Memory leaks may occur when distributed hash table (DHT) updates are triggered by master/replica switchovers or configuration changes. For more information, see Manually switch workloads from a master node to a replica node and Change the configurations of an instance.

• The following issue is fixed: Memory leaks may occur when clients require a long time to receive responses.

• Variables can be used in the indexes of the KEYS array in Lua scripts.

• The following issue is fixed: Memory leaks occur when the MULTI or BLOCK command is run on a cluster instance that has more than 32 data shards. For more information, see Cluster master-replica instances.

• The following issue is fixed: Virtual IP addresses (VIPs) cannot be obtained by using sockets in alb enat mode.

• The IP address of the data shard that sent the most recent response can be recorded when slow logs record the commands used to manage multiple keys.

• The following issue is fixed: The system may not respond when the GIS.GETALL command is run. For more information about this command, see TairGIS.

• The error message returned due to an invalid password is optimized to facilitate your understanding.

• The following issue is fixed: The system does not respond when SSL encryption is enabled in multi-threading mode. For more information, see Configure SSL encryption.
• The following issue is fixed: The response protocol is incorrect due to channels that contain 0 when the UNSUBSCRIBE command is run.

• Operational logs encode and then record binary requests for large packets, acknowledgment packets, and moved packets to prevent garbled characters for logs.

• The issue that the max_session_processing parameter cannot be dynamically configured is fixed. This parameter specifies the maximum number of pending requests that are allowed per connection. For more information about the parameter, see Supported parameters.

• The internal monitoring of proxy nodes is optimized.

• The following issue is fixed: The SLOWLOG command that is run on a standard or cluster instance times out because the node roles are not initialized.
• The following issue is fixed: Connections to ApsaraDB for Memcache instances of specified specifications cannot be established by using Data Management (DMS). For more information, see Overview and What is ApsaraDB for Memcache?
• The following issue is fixed: The system does not respond because no keys are specified when clients subscribe to __keyspace@0__.

• More internal commands of ApsaraDB for Redis are available.

• The special implementation of pipelines on the Jedis client is adapted to optimize the process to disable connections that are beyond the configured connection limit. For more information about Jedis connections, see Jedis client.

• The following issue is fixed: The system incorrectly records slow logs by using the BZPOPMIN and XREAD commands. For more information, see View slow logs.

• The issue that idle connections are automatically closed after 1 minute is fixed.

• Idle connections are automatically closed after a timeout configuration is added.
• Statistics of slow logs on read replicas can be collected. This indicates that the SLOWLOG command is sent to the master node and all read replicas. For more information, see View slow logs.

• The memory usage of connections supported by the PubSub and MONITOR commands is optimized to prevent a rapid increase in memory usage due to memory fragments.
• The capabilities of proxy nodes to handle new connections are improved.

• The following issue is fixed: The system does not respond when you call the stat() function before configurations are complete by using a Configserver node.

• The IP addresses of clients are recorded in slow logs to facilitate troubleshooting. For more information, see View slow logs.

• The capabilities of proxy nodes to handle short-lived connections are improved.

• The following issue is fixed: Memory usage increases because the vector::clear() function does not take effect.

2020-07-13

• The following issue is fixed: The system may not respond when you establish a connection after SSL encryption is enabled.

• Binary data in audit logs is encoded to improve log readability.
• The no_loose_statistics-ip-enable, no_loose_statistics-keys, and no_loose_statistics-cmds parameters are added. You can set the parameters to collect statistics of specific IP addresses, keys, and commands. For more information, see Supported parameters.

• The following issue is fixed: After a connection is closed, the system may not respond when the CheckExceedLimitAndClose command is run.
• The issue that the SSL encryption feature cannot be enabled is fixed.

• The following issue is fixed: The system may not respond to subsequent requests due to empty packets such as \r\n.