monolake-core

monolake-core is a foundational crate for building high-performance, thread-per-core network services. It provides a robust framework for worker orchestration, service deployment, and lifecycle management, supporting protocols such as HTTP and Thrift. This crate builds upon concepts from the service_async crate to implement a thread-per-core worker system with advanced service management capabilities.

Key Features

• Network Service Foundation: Core building blocks for creating efficient network services.
• Thread-per-Core Architecture: Maximizes performance on multi-core processors.
• Service Lifecycle Management: Seamless updates and deployments of service chains.
• Flexible Deployment Models: Support for both single-stage and two-stage service deployment processes.
• State Transfer: Facilitate updates with state preservation between service versions.
• Protocol Support: Built-in support for HTTP and Thrift protocols.
• Asynchronous Design: Leverages Rust's async capabilities for efficient, non-blocking operations.

Service and Service Factory Concepts

This crate builds upon the service_async crate, providing:

• A refined Service trait that leverages impl Trait for improved performance and flexibility.
• The AsyncMakeService trait for efficient creation and updating of services, particularly useful for managing stateful resources across service updates.

monolake-core extends these concepts to provide a comprehensive system for managing network services in a thread-per-core architecture.

Pre-built Services

While monolake-core provides the foundation, you can find pre-built services for common protocols in the monolake-services crate. This includes ready-to-use implementations for:

• HTTP services
• Thrift services

These pre-built services can be easily integrated into your monolake-core based applications, speeding up development for standard network service scenarios.

Worker-Service Lifecycle Management

The core of this crate is the worker-service lifecycle management system, implemented in the orchestrator module. Key components include:

• WorkerManager: Manages multiple worker threads, each running on a dedicated CPU core.
• ServiceExecutor: Handles the lifecycle of services within a single worker thread.
• ServiceDeploymentContainer: Manages individual service instances, including precommitting and deployment.
• ServiceCommand: Represents actions to be performed on services, such as precommitting, updating, or removing.

This system supports dynamic updating of deployed services:

• You can update a currently deployed service with a new service chain.
• Existing connections continue to use the old service chain.
• New connections automatically use the latest service chain.

This approach ensures smooth transitions during updates with minimal disruption to ongoing operations.

Deployment Models

The system supports two deployment models:

1. Two-Stage Deployment: Ideal for updating services while preserving state.

   • Precommit a service using Precommit.
   • Update using Update or commit using Commit.

2. Single-Stage Deployment: Suitable for initial deployments or when state preservation isn't necessary.

   • Create and deploy in one step using PrepareAndCommit.

Protocol Handlers

HTTP Handler

The http module provides the HttpHandler trait for implementing HTTP request handlers. It supports context-aware handling and connection management.

Thrift Handler

The thrift module offers the ThriftHandler trait for implementing Thrift request handlers.

Both handler traits are automatically implemented for types that implement the Service trait with appropriate request and response types.

Usage Example

   let mut manager = WorkerManager::new(config.runtime);
   let join_handlers = manager.spawn_workers_async();
   for (name, ServiceConfig { listener, server }) in config.servers.into_iter() {
        let lis_fac = ListenerBuilder::try_from(listener).expect("build listener failed");
        let svc_fac = l7_factory(server);
        manager
            .dispatch_service_command(ServiceCommand::PrepareAndCommit(
                   Arc::new(name),
                   AsyncMakeServiceWrapper(svc_fac),
                   AsyncMakeServiceWrapper(Arc::new(lis_fac)),
             ))
             .await
             .err()
             .expect("apply init config failed");
   }  

Modules

• orchestrator: Core functionality for worker management and service deployment.
• http: HTTP-specific implementations and utilities.
• thrift: Thrift protocol support and related functionalities.
• config: Configuration structures and utilities for the system.
• context: Context management for request processing.
• listener: Network listener implementations and abstractions.
• util: Various utility functions and helpers.

Error Handling

This crate uses AnyError as a type alias for anyhow::Error, providing flexible error handling. The AnyResult type alias offers a convenient way to return results that can contain any error type.