Hippo4j Documentation

Hippo4j is a dynamically observable thread pool management framework designed to improve system reliability. It addresses common thread pool pain points such as resource exhaustion, difficult parameter estimation, lack of monitoring, and task accumulation.

Key features include:

• Global Management: Manage application thread pool instances.
• Dynamic Configuration: Change parameters like core/max threads, queue capacity, and rejection policies at runtime.
• Alerting: Built-in notification strategies for thread pool activity, capacity watermarks, rejection policies, and long-running tasks.
• Data Collection: Supports Prometheus, InfluxDB, ElasticSearch, logs, and built-in collection.
• Monitoring: Real-time runtime data visualization.
• Context Propagation: Supports passing thread-local contexts (e.g., MDC) to tasks.
• Graceful Shutdown: Supports waiting for tasks to complete within a specified time during project shutdown.
• Framework/Container Support: Adapts to Tomcat, Jetty, Undertow, Dubbo, Hystrix, RabbitMQ, and RocketMQ.

Hippo4j enhances JDK thread pools and extends third-party framework thread pools to improve the operational reliability of business systems.

Key capabilities include:

• Global Management: Manage application thread pool instances.
• Dynamic Configuration: Change parameters at runtime, including core pool size, maximum pool size, blocking queue capacity, and rejection policies.
• Alerting: Built-in notification strategies for thread pool activity, capacity watermarks, rejection policies, and excessively long task execution times.
• Monitoring: Real-time viewing of runtime data and charts for the last 30 minutes of operation.
• Functional Extensions: Support for task context propagation (e.g., MDC) and graceful shutdown (waiting for tasks to complete within a specified time).
• Deployment Modes: Supports both Configuration Center Dependency mode and Middleware-free mode.
• Container & Framework Support: Monitor and modify thread counts for Tomcat, Jetty, Undertow, Dubbo, Hystrix, RabbitMQ, and RocketMQ.

Hippo4j is a lightweight, dynamic, and observable thread pool management system designed to enhance native Java thread pools. It addresses common production issues such as improper parameter configuration, lack of monitoring, task accumulation, and loss of context (e.g., MDC) during asynchronous execution.

Key Problems Solved:

• Resource Management: Prevents high server load caused by excessive thread resources.
• Parameter Tuning: Enables dynamic adjustment of thread pool parameters to match business concurrency.
• Observability: Provides visibility into task execution times, task accumulation, and rejection strategies.
• Context Propagation: Supports passing runtime variables (like MDC context) to threads.
• Graceful Shutdown: Ensures running tasks are not discarded when the application shuts down.
• Deadlock Detection: Helps identify tasks that stop executing due to deadlocks or long-running operations.

Management Hierarchy:

Hippo4j organizes thread pools using a hierarchical structure: Tenant $\rightarrow$ Project $\rightarrow$ Thread Pool. This allows for granular control and permission management across different teams and services.

You can modify dynamic thread pool parameters through the Hippo4j Server Console.

1. Navigate to the Dynamic Thread Pool (动态线程池) menu in the console.
2. Select the specific thread pool instance you wish to modify.
3. Update the parameters and apply changes.

When deploying a Client cluster, you can choose to modify a single specific instance or use the 全部修改 (Modify All) button to apply changes to all instances in the cluster simultaneously.

Successful configuration changes will be reflected in the application logs. A successful change log entry will look similar to this:

2022-09-10 00:23:29.783  INFO 50322 --- [change.config_0] c.h.s.s.c.ServerThreadPoolDynamicRefresh : [message-consume] Dynamic thread pool change parameter.
    corePoolSize: 2 => 4
    maximumPoolSize: 6 => 12
    capacity: 1024 => 2048
    keepAliveTime: 9999 => 9999
    executeTimeOut: 800 => 3000
    rejectedType: SyncPutQueuePolicy => RunsOldestTaskPolicy
    allowCoreThreadTimeOut: true => true