tari-project / tari / 15757511687

62%

Build Type

push

github

Committed by web-flow

Commit Message

fix: peer dialling (#7218)

## Description

Implements proactive peer dialing functionality to resolve connectivity
issues where nodes consistently dial fewer peers than needed to reach
target connection counts. The system now automatically dials 2-3x the
target number of peers to account for connection failures and ensures
robust network connectivity.

**Key Components:**
- **Proactive Dialer**: Core service that calculates optimal dial counts
based on historical success rates and intelligently selects healthy peer
candidates
- **Enhanced Connection Stats**: Extended `PeerConnectionStats` with
circuit breaker pattern integration and health scoring capabilities
- **Peer Health Metrics**: Circuit breaker pattern with time-windowed
success rate tracking and health scoring for peer selection
- **Health-Aware Peer Selection**: Enhanced `ConnectivitySelection` with
health-based prioritization for peer connection choices
- **Peer Discovery Bridge**: Triggers additional peer discovery when
insufficient candidates are available
- **Comprehensive Metrics**: Prometheus metrics for monitoring dial
attempts, success rates, circuit breaker states, and performance

**Architecture:**
- Actor-based service pattern with async message passing
- Integrates into existing ConnectivityManager's connection pool refresh
cycle (currently 180 seconds/3 minutes)
- Configurable multipliers and thresholds for different network
conditions
- Backward compatible with existing connectivity management

**Production Validation:** This implementation has been tested and
verified working on a live mainnet node, with logs confirming proper
execution and logic flow.

---

## Motivation and Context

**Problem:** Nodes were consistently dialing only ~5 peers instead of
sufficient peers to reach the target of 8 connections, leading to poor
network connectivity and potential isolation issues.

**Root Cause:** The existing connectivity manager only dialed exactly
the number of missing connectio... (continued)

Run Details

522 of 760 new or added lines in 10 files covered. (68.68%)

72 existing lines in 14 files now uncovered.

82790 of 114637 relevant lines covered (72.22%)

240690.26 hits per line