Update Type | Update | Detailed Description |
Optimization | Engine version upgrade | The engine is upgraded to the community stable version, addressing critical issues such as graceful restart of Broker, upgrade of Bookie version, optimized memory usage for delayed messages, and optimized caching policy in the computing layer. |
New feature | Topic policies | Topic policies are comprehensively supported. You can flexibly configure policies across multiple dimensions. |
New feature | Monitoring alarms | New consumption monitoring metrics are supported on the backend, enhancing cluster monitoring capabilities. |
Optimization | Performance optimization | Memory usage for delayed messages is optimized. Caching policy in the computing layer is optimized. The load balancing policy and Bundle unloading mechanism are optimized. The storage layer supports DirectIO, enhancing read-write performance at the storage layer. The performance of the management API is optimized. |
Update Type | Update | Detailed Description |
Optimization | Engine version upgrade | The engine is upgraded to the community stable version, addressing critical issues such as graceful restart of Broker, upgrade of Bookie version, optimization of the Key_Shared subscription mode, and inability to scale in unack map. |
New feature | Topic TTL | Topic policies are introduced. You can flexibly configure finer-grained time-to-live (TTL) for messages. This feature is under productization. |
New feature | Consumption monitoring alarm | New consumption monitoring metrics are supported on the backend, enhancing cluster monitoring capabilities. |
Optimization | Performance optimization | The sequentiality guarantee in the Key_Shared subscription mode is optimized. The processing performance of the server under a large-scale message backlog is improved. The speed of disaster recovery for availability zones is improved. The caching policy is optimized to enhance the consumption performance. |
Update Type | Update | Detailed Description | Documentation |
Optimization | Unified domain name access | A unique domain name is automatically generated for each network environment of each cluster as the access address. You do not need to manually create and manage access points. The dependency that declares listenerName is removed. The SDK is more compatible with TCP clients in the community that are not adapted to multiple advertised listeners in Pulsar (including C++, Python, and Node.js). | |
New feature | Support for public network access | Public network access is supported, so you can access a TDMQ for Apache Pulsar cluster in Tencent Cloud from a TCP client in your local systems or self-built Internet Data Centers (IDCs). | |
Optimization | Update of the naming rules for the retry and dead letter topics | The naming rules for the retry and dead letter topics are updated to the same as used by the Pulsar community, so that they are more compatible with open-source SDKs (the new naming rules are not applicable to the original Tencent Cloud SDK). | |
Optimization | Performance optimization | Clusters of version 2.7.1 are optimized at the system kernel level to achieve an approximately 20% higher peak production and consumption performance than clusters of version 2.6.x. In addition, they can be scaled out imperceptibly based on actual usage patterns. | - |
Update Type | Update | Detailed Description | Documentation |
Optimization | Stability improvement | Issues discovered during the beta test are fixed, significantly improving the cluster stability. | - |
Optimization | ZooKeeper hot and cold isolation | The ZooKeeper component of the Pulsar kernel adopts the deployment architecture of hot and cold data isolation, where metadata and cluster broker configuration data are deployed separately in two ZooKeeper instances to enhance cluster stability. | - |
New feature | Support for the creation of virtual clusters | You can create virtual clusters in the console for resource isolation. | - |
フィードバック