[Enhancement]: Global GC Manager for morpc - Align with gRPC/Thrift Best Practices

[XuPeng-SH]

Is there an existing issue for the same feature request?

• I have checked the existing issues.

Is your feature request related to a problem?

Currently, each morpc client creates 3 goroutines for GC tasks. With 104 clients in TN, this results in 312 goroutines, contributing to high goroutine count (1011 total).

Describe the feature you'd like

Implement a global GC manager following the same design pattern as mainstream RPC frameworks (gRPC, Thrift, Dubbo):

Current Design (Per-Client):

• Each client: 3 goroutines (gcIdleTask, gcInactiveTask, createTask)
• 104 clients → 312 goroutines

Proposed Design (Global Manager - Like gRPC):

• Global GC manager: 3 shared goroutines
• All clients register with global manager
• Reduces goroutines: 312 → 3 (99% reduction)

Architecture Alignment with Mainstream Frameworks

gRPC Model

• Global Transport layer manages all connections
• Per-connection: 2 goroutines (read/write loops)
• Connection-level GC: Unified management

Thrift Model

• Global connection pool manager
• Per-connection: 2 goroutines (read/write loops)
• Connection-level GC: Unified management

morpc New Design (Proposed)

• Global GC manager (similar to gRPC Transport layer)
• Per-backend: 2 goroutines (read/write loops)
• Connection-level GC: Unified management via global manager

Implementation Details

1. Global GC Manager (pkg/common/morpc/client_gc.go)

   • Singleton pattern (like gRPC's Transport layer)
   • 3 shared goroutines for all clients
   • Thread-safe registration/unregistration
   • Copy-on-Write pattern for minimal lock contention

2. Client Integration (pkg/common/morpc/client.go)

   • Remove per-client goroutines
   • Register with global manager on creation
   • Unregister on close

3. Configuration

   • Environment: MORPC_GC_IDLE_INTERVAL, MORPC_GC_CHANNEL_BUFFER_SIZE
   • Config file: GCIdleCheckInterval, GCChannelBufferSize
   • Defaults: 1 second check interval, 4096 channel buffer

4. Monitoring

   • Prometheus metrics: mo_rpc_gc_channel_drop_total
   • Grafana dashboard integration

Benefits

• Aligns with gRPC, Thrift, Dubbo best practices
• Reduces goroutines: 312 → 3 (99% reduction)
• Reduces TN goroutines: 1011 → ~700 (30% reduction)
• Maintains backward compatibility
• Improves resource efficiency

Testing

• 16 unit tests covering all GC manager functionality
• Stress tests with 50 clients × 100 operations
• Race condition detection with -race flag
• All tests passing

Performance Metrics

• Goroutine reduction: 312 → 3 (99%)
• Lock contention: O(N) → O(1) via Copy-on-Write
• Thundering herd prevention: ±10% Jitter

Aligns with industry best practices