unrealon 1.0.9__py3-none-any.whl → 1.1.0__py3-none-any.whl

unrealon_sdk/src/enterprise/load_balancer.py

@@ -1,964 +0,0 @@
-"""
-Load Balancer - Layer 4 Concurrency Service
-
-Enterprise-grade load balancing system with intelligent traffic distribution,
-health monitoring, and adaptive algorithms. Provides optimal resource utilization
-for parsing operations with geographic awareness and performance optimization.
-
-Features:
-- Multiple load balancing algorithms (round-robin, least connections, performance-based)
-- Dynamic node health monitoring with circuit breaker patterns
-- Geographic and proximity-based routing for parsing operations
-- Session affinity and sticky sessions support
-- Real-time performance metrics and adaptive optimization
-- Failover strategies with graceful degradation
-- Integration with proxy management and resource pooling
-- Configurable traffic routing rules and policies
-- Connection draining and maintenance mode support
-- Advanced analytics and decision tracking
-"""
-
-import asyncio
-import logging
-import time
-import threading
-import hashlib
-import random
-from typing import Dict, List, Optional, Any, Set, Callable, Union
-from datetime import datetime, timezone, timedelta
-from collections import defaultdict, deque
-from dataclasses import dataclass, field
-import weakref
-
-# Core SDK components
-from unrealon_sdk.src.core.config import AdapterConfig
-from unrealon_sdk.src.utils import generate_correlation_id
-
-# DTO models
-from unrealon_sdk.src.dto.logging import SDKEventType, SDKSeverity
-from unrealon_sdk.src.dto.concurrency import ConcurrencyEventType, ConcurrencyMetrics
-from unrealon_sdk.src.dto.load_balancing import (
-    LoadBalancingAlgorithm,
-    NodeHealthStatus,
-    TrafficDirection,
-    FailoverStrategy,
-    CircuitBreakerState,
-    LoadBalancerNode,
-    LoadBalancingRule,
-    LoadBalancingDecisionRequest,
-    LoadBalancingDecisionResult,
-    LoadBalancerStatistics,
-    HealthCheckConfig,
-    LoadBalancingSession,
-)
-
-# Development logging
-from typing import TYPE_CHECKING
-
-if TYPE_CHECKING:
-    from unrealon_sdk.src.enterprise.logging import DevelopmentLogger
-
-logger = logging.getLogger(__name__)
-
-
-@dataclass
-class NodePerformanceTracker:
-    """Track node performance metrics."""
-
-    node_id: str
-    response_times: deque[float] = field(default_factory=lambda: deque(maxlen=100))
-    success_count: int = 0
-    failure_count: int = 0
-    last_request: Optional[datetime] = None
-    avg_response_time: float = 0.0
-    success_rate: float = 1.0
-    load_score: float = 0.0
-
-
-class LoadBalancer:
-    """
-    Enterprise-grade load balancer.
-
-    Provides intelligent traffic distribution with health monitoring,
-    performance optimization, and adaptive algorithms for parsing operations.
-    """
-
-    def __init__(
-        self,
-        config: AdapterConfig,
-        health_check_config: Optional[HealthCheckConfig] = None,
-        dev_logger: Optional["DevelopmentLogger"] = None,
-    ):
-        """Initialize load balancer."""
-        self.config = config
-        self.health_check_config = health_check_config or HealthCheckConfig()
-        self.dev_logger = dev_logger
-
-        # Node management
-        self._nodes: Dict[str, LoadBalancerNode] = {}
-        self._node_performance: Dict[str, NodePerformanceTracker] = {}
-        self._healthy_nodes: Set[str] = set()
-        self._unhealthy_nodes: Set[str] = set()
-
-        # Load balancing rules
-        self._rules: Dict[str, LoadBalancingRule] = {}
-        self._default_rule: Optional[LoadBalancingRule] = None
-
-        # Algorithm implementations
-        self._algorithms: Dict[LoadBalancingAlgorithm, Callable] = {
-            LoadBalancingAlgorithm.ROUND_ROBIN: self._round_robin_selection,
-            LoadBalancingAlgorithm.WEIGHTED_ROUND_ROBIN: self._weighted_round_robin_selection,
-            LoadBalancingAlgorithm.LEAST_CONNECTIONS: self._least_connections_selection,
-            LoadBalancingAlgorithm.WEIGHTED_LEAST_CONNECTIONS: self._weighted_least_connections_selection,
-            LoadBalancingAlgorithm.LEAST_RESPONSE_TIME: self._least_response_time_selection,
-            LoadBalancingAlgorithm.RESOURCE_BASED: self._resource_based_selection,
-            LoadBalancingAlgorithm.GEOGRAPHIC: self._geographic_selection,
-            LoadBalancingAlgorithm.HASH_BASED: self._hash_based_selection,
-            LoadBalancingAlgorithm.RANDOM: self._random_selection,
-            LoadBalancingAlgorithm.ADAPTIVE: self._adaptive_selection,
-        }
-
-        # Session management
-        self._sessions: Dict[str, LoadBalancingSession] = {}
-        self._session_cleanup_threshold = timedelta(hours=24)
-
-        # State tracking
-        self._round_robin_index: Dict[str, int] = defaultdict(int)
-        self._decision_cache: Dict[str, LoadBalancingDecisionResult] = {}
-        self._cache_ttl_seconds = 60.0
-
-        # Statistics
-        self._statistics = LoadBalancerStatistics()
-        self._decision_times: deque[float] = deque(maxlen=1000)
-
-        # Background tasks
-        self._health_check_task: Optional[asyncio.Task[None]] = None
-        self._cleanup_task: Optional[asyncio.Task[None]] = None
-        self._metrics_task: Optional[asyncio.Task[None]] = None
-        self._shutdown = False
-
-        # Thread safety
-        self._lock = threading.RLock()
-
-        self._log_info("Load balancer initialized")
-
-    async def start(self) -> None:
-        """Start load balancer."""
-        # Start background tasks
-        if self._health_check_task is None:
-            self._health_check_task = asyncio.create_task(self._health_check_loop())
-
-        if self._cleanup_task is None:
-            self._cleanup_task = asyncio.create_task(self._cleanup_loop())
-
-        if self._metrics_task is None:
-            self._metrics_task = asyncio.create_task(self._metrics_loop())
-
-        self._log_info("Load balancer started")
-
-    async def stop(self) -> None:
-        """Stop load balancer."""
-        self._shutdown = True
-
-        # Cancel background tasks
-        for task in [self._health_check_task, self._cleanup_task, self._metrics_task]:
-            if task:
-                task.cancel()
-                try:
-                    await task
-                except asyncio.CancelledError:
-                    pass
-
-        self._log_info("Load balancer stopped")
-
-    def add_node(self, node: LoadBalancerNode) -> None:
-        """Add node to load balancer."""
-        with self._lock:
-            self._nodes[node.node_id] = node
-            self._node_performance[node.node_id] = NodePerformanceTracker(node_id=node.node_id)
-
-            if node.is_enabled and node.health_status == NodeHealthStatus.HEALTHY:
-                self._healthy_nodes.add(node.node_id)
-            else:
-                self._unhealthy_nodes.add(node.node_id)
-
-        self._log_info(f"Added node '{node.node_name}' ({node.host}:{node.port})")
-
-    def remove_node(self, node_id: str) -> bool:
-        """Remove node from load balancer."""
-        with self._lock:
-            if node_id not in self._nodes:
-                return False
-
-            del self._nodes[node_id]
-            del self._node_performance[node_id]
-            self._healthy_nodes.discard(node_id)
-            self._unhealthy_nodes.discard(node_id)
-
-        self._log_info(f"Removed node '{node_id}'")
-        return True
-
-    def add_rule(self, rule: LoadBalancingRule) -> None:
-        """Add load balancing rule."""
-        with self._lock:
-            self._rules[rule.rule_id] = rule
-
-        self._log_info(f"Added load balancing rule '{rule.rule_name}'")
-
-    def set_default_rule(self, rule: LoadBalancingRule) -> None:
-        """Set default load balancing rule."""
-        self._default_rule = rule
-        self._log_info(f"Set default rule to '{rule.rule_name}'")
-
-    async def select_node(
-        self, request: LoadBalancingDecisionRequest
-    ) -> LoadBalancingDecisionResult:
-        """Select node for request using load balancing."""
-        start_time = time.time()
-
-        try:
-            # Check cache first
-            cache_key = self._generate_cache_key(request)
-            if cache_key in self._decision_cache:
-                cached_result = self._decision_cache[cache_key]
-                if self._is_cache_valid(cached_result):
-                    return cached_result
-
-            # Find applicable rule
-            rule = self._find_applicable_rule(request)
-            if not rule:
-                rule = self._default_rule
-
-            if not rule:
-                raise RuntimeError("No load balancing rule available")
-
-            # Check session affinity
-            if rule.session_affinity and request.session_id:
-                session_result = self._handle_session_affinity(request, rule)
-                if session_result:
-                    return session_result
-
-            # Get candidate nodes
-            candidate_nodes = self._get_candidate_nodes(request, rule)
-            if not candidate_nodes:
-                raise RuntimeError("No healthy nodes available")
-
-            # Apply load balancing algorithm
-            selected_node = self._apply_algorithm(rule.algorithm, candidate_nodes, request)
-            if not selected_node:
-                raise RuntimeError("Load balancing algorithm failed to select node")
-
-            # Get backup nodes
-            backup_nodes = self._get_backup_nodes(selected_node, candidate_nodes, rule)
-
-            # Create decision result
-            decision_time = (time.time() - start_time) * 1000
-            result = LoadBalancingDecisionResult(
-                selected_node=selected_node,
-                backup_nodes=backup_nodes,
-                algorithm_used=rule.algorithm,
-                rule_applied=rule.rule_id,
-                decision_time_ms=decision_time,
-                selection_factors=self._get_selection_factors(selected_node, rule.algorithm),
-                timestamp=datetime.now(timezone.utc),
-            )
-
-            # Handle session creation
-            if rule.session_affinity and request.session_id:
-                self._create_session(request, selected_node)
-                result.new_session_created = True
-
-            # Cache decision
-            self._decision_cache[cache_key] = result
-
-            # Update statistics
-            self._update_statistics(result)
-
-            return result
-
-        except Exception as e:
-            decision_time = (time.time() - start_time) * 1000
-            self._statistics.failed_requests += 1
-
-            error_result = LoadBalancingDecisionResult(
-                selected_node=None,
-                algorithm_used=(
-                    rule.algorithm if "rule" in locals() else LoadBalancingAlgorithm.ROUND_ROBIN
-                ),
-                decision_time_ms=decision_time,
-                rejection_reasons={"error": str(e)},
-                timestamp=datetime.now(timezone.utc),
-            )
-
-            self._log_error(f"Load balancing failed: {e}")
-            return error_result
-
-    def report_request_result(
-        self,
-        node_id: str,
-        success: bool,
-        response_time_ms: float,
-        error_message: Optional[str] = None,
-    ) -> None:
-        """Report request result for node performance tracking."""
-        with self._lock:
-            if node_id not in self._nodes:
-                return
-
-            node = self._nodes[node_id]
-            performance = self._node_performance[node_id]
-
-            # Update performance metrics
-            performance.response_times.append(response_time_ms)
-            performance.last_request = datetime.now(timezone.utc)
-
-            if success:
-                performance.success_count += 1
-                node.total_requests += 1
-            else:
-                performance.failure_count += 1
-                node.failed_requests += 1
-                node.failure_count += 1
-
-            # Calculate averages
-            if performance.response_times:
-                performance.avg_response_time = sum(performance.response_times) / len(
-                    performance.response_times
-                )
-                node.avg_response_time_ms = performance.avg_response_time
-
-            total_requests = performance.success_count + performance.failure_count
-            if total_requests > 0:
-                performance.success_rate = performance.success_count / total_requests
-
-            # Update circuit breaker state
-            self._update_circuit_breaker(node, success, error_message)
-
-            # Update node health
-            self._update_node_health(node_id, success)
-
-    def _find_applicable_rule(
-        self, request: LoadBalancingDecisionRequest
-    ) -> Optional[LoadBalancingRule]:
-        """Find applicable load balancing rule for request."""
-        for rule in sorted(self._rules.values(), key=lambda r: r.priority):
-            if self._rule_matches(rule, request):
-                return rule
-        return None
-
-    def _rule_matches(self, rule: LoadBalancingRule, request: LoadBalancingDecisionRequest) -> bool:
-        """Check if rule matches request."""
-        # Check traffic direction
-        if rule.traffic_direction != TrafficDirection.BIDIRECTIONAL:
-            if rule.traffic_direction != request.traffic_direction:
-                return False
-
-        # Check source patterns
-        if rule.source_patterns:
-            if not any(
-                self._pattern_matches(pattern, request.source_ip)
-                for pattern in rule.source_patterns
-            ):
-                return False
-
-        # Check path patterns (if HTTP)
-        if rule.path_patterns and request.path:
-            if not any(
-                self._pattern_matches(pattern, request.path) for pattern in rule.path_patterns
-            ):
-                return False
-
-        return True
-
-    def _pattern_matches(self, pattern: str, value: str) -> bool:
-        """Check if pattern matches value (simple wildcard support)."""
-        if "*" not in pattern:
-            return pattern == value
-
-        # Simple wildcard matching
-        parts = pattern.split("*")
-        if not value.startswith(parts[0]):
-            return False
-        if not value.endswith(parts[-1]):
-            return False
-
-        return True
-
-    def _get_candidate_nodes(
-        self, request: LoadBalancingDecisionRequest, rule: LoadBalancingRule
-    ) -> List[LoadBalancerNode]:
-        """Get candidate nodes for request."""
-        candidates = []
-
-        # Start with rule target nodes or all healthy nodes
-        target_node_ids = rule.target_nodes if rule.target_nodes else list(self._healthy_nodes)
-
-        for node_id in target_node_ids:
-            if node_id not in self._nodes:
-                continue
-
-            node = self._nodes[node_id]
-
-            # Check if node is enabled and healthy
-            if not node.is_enabled or node.health_status not in [
-                NodeHealthStatus.HEALTHY,
-                NodeHealthStatus.DEGRADED,
-            ]:
-                continue
-
-            # Check circuit breaker
-            if node.circuit_breaker_state == CircuitBreakerState.OPEN:
-                continue
-
-            # Check required tags
-            if request.required_tags:
-                if not all(node.tags.get(k) == v for k, v in request.required_tags.items()):
-                    continue
-
-            # Check excluded nodes
-            if node.node_id in request.excluded_nodes:
-                continue
-
-            # Check connection limits
-            if node.max_connections and node.current_connections >= node.max_connections:
-                continue
-
-            candidates.append(node)
-
-        return candidates
-
-    def _apply_algorithm(
-        self,
-        algorithm: LoadBalancingAlgorithm,
-        candidates: List[LoadBalancerNode],
-        request: LoadBalancingDecisionRequest,
-    ) -> Optional[LoadBalancerNode]:
-        """Apply load balancing algorithm to select node."""
-        if not candidates:
-            return None
-
-        algorithm_func = self._algorithms.get(algorithm)
-        if not algorithm_func:
-            # Fallback to round robin
-            algorithm_func = self._algorithms[LoadBalancingAlgorithm.ROUND_ROBIN]
-
-        return algorithm_func(candidates, request)
-
-    def _round_robin_selection(
-        self, candidates: List[LoadBalancerNode], request: LoadBalancingDecisionRequest
-    ) -> Optional[LoadBalancerNode]:
-        """Round-robin selection algorithm."""
-        if not candidates:
-            return None
-
-        rule_key = request.request_id[:8]  # Use part of request ID as key
-        index = self._round_robin_index[rule_key] % len(candidates)
-        self._round_robin_index[rule_key] = (index + 1) % len(candidates)
-
-        return candidates[index]
-
-    def _weighted_round_robin_selection(
-        self, candidates: List[LoadBalancerNode], request: LoadBalancingDecisionRequest
-    ) -> Optional[LoadBalancerNode]:
-        """Weighted round-robin selection algorithm."""
-        if not candidates:
-            return None
-
-        # Create weighted list
-        weighted_candidates = []
-        for node in candidates:
-            weighted_candidates.extend([node] * node.weight)
-
-        if not weighted_candidates:
-            return candidates[0]
-
-        rule_key = f"weighted_{request.request_id[:8]}"
-        index = self._round_robin_index[rule_key] % len(weighted_candidates)
-        self._round_robin_index[rule_key] = (index + 1) % len(weighted_candidates)
-
-        return weighted_candidates[index]
-
-    def _least_connections_selection(
-        self, candidates: List[LoadBalancerNode], request: LoadBalancingDecisionRequest
-    ) -> Optional[LoadBalancerNode]:
-        """Least connections selection algorithm."""
-        if not candidates:
-            return None
-
-        return min(candidates, key=lambda node: node.current_connections)
-
-    def _weighted_least_connections_selection(
-        self, candidates: List[LoadBalancerNode], request: LoadBalancingDecisionRequest
-    ) -> Optional[LoadBalancerNode]:
-        """Weighted least connections selection algorithm."""
-        if not candidates:
-            return None
-
-        # Calculate weighted connection ratio
-        def weighted_ratio(node: LoadBalancerNode) -> float:
-            if node.weight <= 0:
-                return float("inf")
-            return node.current_connections / node.weight
-
-        return min(candidates, key=weighted_ratio)
-
-    def _least_response_time_selection(
-        self, candidates: List[LoadBalancerNode], request: LoadBalancingDecisionRequest
-    ) -> Optional[LoadBalancerNode]:
-        """Least response time selection algorithm."""
-        if not candidates:
-            return None
-
-        return min(candidates, key=lambda node: node.avg_response_time_ms)
-
-    def _resource_based_selection(
-        self, candidates: List[LoadBalancerNode], request: LoadBalancingDecisionRequest
-    ) -> Optional[LoadBalancerNode]:
-        """Resource-based selection algorithm."""
-        if not candidates:
-            return None
-
-        # Calculate resource utilization score
-        def resource_score(node: LoadBalancerNode) -> float:
-            connection_ratio = 0.0
-            if node.max_connections:
-                connection_ratio = node.current_connections / node.max_connections
-
-            # Consider response time and failure rate
-            performance = self._node_performance.get(node.node_id)
-            if performance:
-                response_factor = min(
-                    performance.avg_response_time / 1000.0, 1.0
-                )  # Normalize to 0-1
-                failure_factor = 1.0 - performance.success_rate
-                return connection_ratio * 0.4 + response_factor * 0.3 + failure_factor * 0.3
-
-            return connection_ratio
-
-        return min(candidates, key=resource_score)
-
-    def _geographic_selection(
-        self, candidates: List[LoadBalancerNode], request: LoadBalancingDecisionRequest
-    ) -> Optional[LoadBalancerNode]:
-        """Geographic selection algorithm."""
-        if not candidates:
-            return None
-
-        # If preferred region specified, try to use it
-        if request.preferred_region:
-            regional_candidates = [
-                node for node in candidates if node.region == request.preferred_region
-            ]
-            if regional_candidates:
-                # Use least connections within preferred region
-                return self._least_connections_selection(regional_candidates, request)
-
-        # Fallback to least connections
-        return self._least_connections_selection(candidates, request)
-
-    def _hash_based_selection(
-        self, candidates: List[LoadBalancerNode], request: LoadBalancingDecisionRequest
-    ) -> Optional[LoadBalancerNode]:
-        """Hash-based selection algorithm (consistent hashing)."""
-        if not candidates:
-            return None
-
-        # Use session ID or source IP for hashing
-        hash_key = request.session_id or request.source_ip
-        hash_value = int(hashlib.md5(hash_key.encode(), usedforsecurity=False).hexdigest(), 16)
-
-        return candidates[hash_value % len(candidates)]
-
-    def _random_selection(
-        self, candidates: List[LoadBalancerNode], request: LoadBalancingDecisionRequest
-    ) -> Optional[LoadBalancerNode]:
-        """Random selection algorithm."""
-        if not candidates:
-            return None
-
-        return random.choice(candidates)
-
-    def _adaptive_selection(
-        self, candidates: List[LoadBalancerNode], request: LoadBalancingDecisionRequest
-    ) -> Optional[LoadBalancerNode]:
-        """Adaptive selection algorithm based on current performance."""
-        if not candidates:
-            return None
-
-        # Calculate adaptive score based on multiple factors
-        def adaptive_score(node: LoadBalancerNode) -> float:
-            performance = self._node_performance.get(node.node_id)
-            if not performance:
-                return 0.5  # Neutral score for new nodes
-
-            # Combine multiple metrics
-            response_factor = min(performance.avg_response_time / 1000.0, 1.0)
-            success_factor = performance.success_rate
-            connection_factor = 0.0
-
-            if node.max_connections:
-                connection_factor = 1.0 - (node.current_connections / node.max_connections)
-
-            # Weight factors: success rate is most important, then response time, then connections
-            score = success_factor * 0.5 + (1.0 - response_factor) * 0.3 + connection_factor * 0.2
-            return score
-
-        # Select node with highest adaptive score
-        return max(candidates, key=adaptive_score)
-
-    def _get_backup_nodes(
-        self,
-        selected_node: LoadBalancerNode,
-        candidates: List[LoadBalancerNode],
-        rule: LoadBalancingRule,
-    ) -> List[LoadBalancerNode]:
-        """Get backup nodes for failover."""
-        backup_nodes = []
-
-        # Add explicitly configured backup nodes
-        for backup_id in rule.backup_nodes:
-            if backup_id in self._nodes and backup_id != selected_node.node_id:
-                backup_node = self._nodes[backup_id]
-                if backup_node.is_enabled and backup_node.health_status in [
-                    NodeHealthStatus.HEALTHY,
-                    NodeHealthStatus.DEGRADED,
-                ]:
-                    backup_nodes.append(backup_node)
-
-        # Add other healthy candidates as additional backups
-        for candidate in candidates:
-            if candidate.node_id != selected_node.node_id and candidate not in backup_nodes:
-                backup_nodes.append(candidate)
-                if len(backup_nodes) >= 3:  # Limit backup nodes
-                    break
-
-        return backup_nodes
-
-    def _handle_session_affinity(
-        self, request: LoadBalancingDecisionRequest, rule: LoadBalancingRule
-    ) -> Optional[LoadBalancingDecisionResult]:
-        """Handle session affinity for sticky sessions."""
-        if not request.session_id:
-            return None
-
-        session = self._sessions.get(request.session_id)
-        if not session or not session.is_active:
-            return None
-
-        # Check if assigned node is still healthy
-        if session.assigned_node_id in self._healthy_nodes:
-            assigned_node = self._nodes[session.assigned_node_id]
-
-            # Update session
-            session.last_accessed = datetime.now(timezone.utc)
-            session.request_count += 1
-
-            return LoadBalancingDecisionResult(
-                selected_node=assigned_node,
-                algorithm_used=rule.algorithm,
-                rule_applied=rule.rule_id,
-                decision_time_ms=1.0,  # Cached decision
-                session_affinity_used=True,
-                timestamp=datetime.now(timezone.utc),
-            )
-
-        # Session node is unhealthy, remove session
-        del self._sessions[request.session_id]
-        return None
-
-    def _create_session(
-        self, request: LoadBalancingDecisionRequest, selected_node: LoadBalancerNode
-    ) -> None:
-        """Create new session for sticky sessions."""
-        if not request.session_id:
-            return
-
-        session = LoadBalancingSession(
-            session_id=request.session_id,
-            client_ip=request.source_ip,
-            assigned_node_id=selected_node.node_id,
-            created_at=datetime.now(timezone.utc),
-            last_accessed=datetime.now(timezone.utc),
-        )
-
-        self._sessions[request.session_id] = session
-
-    def _update_circuit_breaker(
-        self, node: LoadBalancerNode, success: bool, error_message: Optional[str]
-    ) -> None:
-        """Update circuit breaker state for node."""
-        if not success:
-            node.failure_count += 1
-            node.last_failure = datetime.now(timezone.utc)
-
-            # Trip circuit breaker if failure threshold reached
-            if node.failure_count >= 5:  # Configurable threshold
-                node.circuit_breaker_state = CircuitBreakerState.OPEN
-                self._log_info(f"Circuit breaker opened for node {node.node_id}")
-        else:
-            # Reset failure count on success
-            if node.circuit_breaker_state == CircuitBreakerState.HALF_OPEN:
-                node.circuit_breaker_state = CircuitBreakerState.CLOSED
-                node.failure_count = 0
-                self._log_info(f"Circuit breaker closed for node {node.node_id}")
-
-    def _update_node_health(self, node_id: str, success: bool) -> None:
-        """Update node health status."""
-        with self._lock:
-            if node_id not in self._nodes:
-                return
-
-            node = self._nodes[node_id]
-            performance = self._node_performance[node_id]
-
-            # Update health based on success rate
-            if performance.success_rate < 0.5:  # Less than 50% success
-                if node.health_status == NodeHealthStatus.HEALTHY:
-                    node.health_status = NodeHealthStatus.DEGRADED
-                    self._log_info(f"Node {node_id} health degraded")
-            elif performance.success_rate > 0.8:  # More than 80% success
-                if node.health_status == NodeHealthStatus.DEGRADED:
-                    node.health_status = NodeHealthStatus.HEALTHY
-                    self._log_info(f"Node {node_id} health recovered")
-
-            # Update healthy/unhealthy sets
-            if (
-                node.health_status in [NodeHealthStatus.HEALTHY, NodeHealthStatus.DEGRADED]
-                and node.is_enabled
-            ):
-                self._healthy_nodes.add(node_id)
-                self._unhealthy_nodes.discard(node_id)
-            else:
-                self._unhealthy_nodes.add(node_id)
-                self._healthy_nodes.discard(node_id)
-
-    async def _health_check_loop(self) -> None:
-        """Background health check loop."""
-        while not self._shutdown:
-            try:
-                await asyncio.sleep(self.health_check_config.interval_seconds)
-                await self._perform_health_checks()
-            except asyncio.CancelledError:
-                break
-            except Exception as e:
-                logger.error(f"Error in health check loop: {e}")
-
-    async def _perform_health_checks(self) -> None:
-        """Perform health checks on all nodes."""
-        if not self.health_check_config.enabled:
-            return
-
-        for node_id, node in list(self._nodes.items()):
-            try:
-                is_healthy = await self._check_node_health(node)
-                self._update_node_health_status(node_id, is_healthy)
-            except Exception as e:
-                logger.error(f"Health check failed for node {node_id}: {e}")
-                self._update_node_health_status(node_id, False)
-
-    async def _check_node_health(self, node: LoadBalancerNode) -> bool:
-        """Check individual node health."""
-        # Simple TCP connection check
-        try:
-            reader, writer = await asyncio.wait_for(
-                asyncio.open_connection(node.host, node.port),
-                timeout=self.health_check_config.timeout_seconds,
-            )
-            writer.close()
-            await writer.wait_closed()
-            return True
-        except Exception:
-            return False
-
-    def _update_node_health_status(self, node_id: str, is_healthy: bool) -> None:
-        """Update node health status based on health check."""
-        with self._lock:
-            if node_id not in self._nodes:
-                return
-
-            node = self._nodes[node_id]
-            node.last_health_check = datetime.now(timezone.utc)
-
-            if is_healthy:
-                if node.health_status == NodeHealthStatus.UNHEALTHY:
-                    node.health_status = NodeHealthStatus.HEALTHY
-                    self._log_info(f"Node {node_id} health recovered")
-            else:
-                if node.health_status in [NodeHealthStatus.HEALTHY, NodeHealthStatus.DEGRADED]:
-                    node.health_status = NodeHealthStatus.UNHEALTHY
-                    self._log_info(f"Node {node_id} marked unhealthy")
-
-            # Update sets
-            if (
-                node.health_status in [NodeHealthStatus.HEALTHY, NodeHealthStatus.DEGRADED]
-                and node.is_enabled
-            ):
-                self._healthy_nodes.add(node_id)
-                self._unhealthy_nodes.discard(node_id)
-            else:
-                self._unhealthy_nodes.add(node_id)
-                self._healthy_nodes.discard(node_id)
-
-    async def _cleanup_loop(self) -> None:
-        """Background cleanup loop."""
-        while not self._shutdown:
-            try:
-                await asyncio.sleep(300)  # Cleanup every 5 minutes
-                await self._cleanup_sessions()
-                await self._cleanup_cache()
-            except asyncio.CancelledError:
-                break
-            except Exception as e:
-                logger.error(f"Error in cleanup loop: {e}")
-
-    async def _cleanup_sessions(self) -> None:
-        """Clean up expired sessions."""
-        current_time = datetime.now(timezone.utc)
-        expired_sessions = []
-
-        for session_id, session in self._sessions.items():
-            if current_time - session.last_accessed > self._session_cleanup_threshold:
-                expired_sessions.append(session_id)
-
-        for session_id in expired_sessions:
-            del self._sessions[session_id]
-
-        if expired_sessions:
-            self._log_info(f"Cleaned up {len(expired_sessions)} expired sessions")
-
-    async def _cleanup_cache(self) -> None:
-        """Clean up expired cache entries."""
-        current_time = time.time()
-        expired_keys = []
-
-        for cache_key, result in self._decision_cache.items():
-            if current_time - result.timestamp.timestamp() > self._cache_ttl_seconds:
-                expired_keys.append(cache_key)
-
-        for cache_key in expired_keys:
-            del self._decision_cache[cache_key]
-
-    async def _metrics_loop(self) -> None:
-        """Background metrics collection loop."""
-        while not self._shutdown:
-            try:
-                await asyncio.sleep(60)  # Collect every minute
-                await self._collect_metrics()
-            except asyncio.CancelledError:
-                break
-            except Exception as e:
-                logger.error(f"Error in metrics loop: {e}")
-
-    async def _collect_metrics(self) -> None:
-        """Collect load balancer metrics."""
-        with self._lock:
-            self._statistics.total_nodes = len(self._nodes)
-            self._statistics.healthy_nodes = len(self._healthy_nodes)
-            self._statistics.unhealthy_nodes = len(self._unhealthy_nodes)
-
-            # Calculate average decision time
-            if self._decision_times:
-                self._statistics.avg_decision_time_ms = sum(self._decision_times) / len(
-                    self._decision_times
-                )
-
-            # Update algorithm usage
-            for node_id, node in self._nodes.items():
-                self._statistics.node_selection_count[node_id] = node.total_requests
-
-    def get_statistics(self) -> LoadBalancerStatistics:
-        """Get load balancer statistics."""
-        return self._statistics.model_copy()
-
-    def get_node_status(self, node_id: str) -> Optional[LoadBalancerNode]:
-        """Get node status."""
-        return self._nodes.get(node_id)
-
-    def get_all_nodes(self) -> List[LoadBalancerNode]:
-        """Get all nodes."""
-        return list(self._nodes.values())
-
-    def get_healthy_nodes(self) -> List[LoadBalancerNode]:
-        """Get healthy nodes."""
-        return [self._nodes[node_id] for node_id in self._healthy_nodes]
-
-    def _generate_cache_key(self, request: LoadBalancingDecisionRequest) -> str:
-        """Generate cache key for request."""
-        key_parts = [
-            request.source_ip,
-            request.destination_ip or "",
-            str(request.destination_port or ""),
-            request.path or "",
-            str(sorted(request.required_tags.items())),
-            str(sorted(request.excluded_nodes)),
-        ]
-        return hashlib.md5("|".join(key_parts).encode(), usedforsecurity=False).hexdigest()
-
-    def _is_cache_valid(self, result: LoadBalancingDecisionResult) -> bool:
-        """Check if cached result is still valid."""
-        age = time.time() - result.timestamp.timestamp()
-        return age < self._cache_ttl_seconds
-
-    def _get_selection_factors(
-        self, node: LoadBalancerNode, algorithm: LoadBalancingAlgorithm
-    ) -> Dict[str, Any]:
-        """Get factors that influenced node selection."""
-        factors = {
-            "node_id": node.node_id,
-            "algorithm": algorithm.value,
-            "current_connections": node.current_connections,
-            "avg_response_time": node.avg_response_time_ms,
-            "health_status": node.health_status.value,
-            "weight": node.weight,
-        }
-
-        performance = self._node_performance.get(node.node_id)
-        if performance:
-            factors.update(
-                {
-                    "success_rate": performance.success_rate,
-                    "load_score": performance.load_score,
-                }
-            )
-
-        return factors
-
-    def _update_statistics(self, result: LoadBalancingDecisionResult) -> None:
-        """Update statistics with decision result."""
-        self._statistics.total_requests += 1
-        if result.selected_node:
-            self._statistics.successful_requests += 1
-
-        self._decision_times.append(result.decision_time_ms)
-
-        # Update algorithm usage
-        algorithm_key = result.algorithm_used.value
-        self._statistics.algorithm_usage[algorithm_key] = (
-            self._statistics.algorithm_usage.get(algorithm_key, 0) + 1
-        )
-
-    def _log_info(self, message: str, **kwargs: Any) -> None:
-        """Log info message."""
-        if self.dev_logger:
-            self.dev_logger.log_info(
-                SDKEventType.PERFORMANCE_OPTIMIZATION_APPLIED, message, **kwargs
-            )
-        else:
-            logger.info(message)
-
-    def _log_error(self, message: str, **kwargs: Any) -> None:
-        """Log error message."""
-        if self.dev_logger:
-            self.dev_logger.log_error(SDKEventType.CRITICAL_ERROR, message, **kwargs)
-        else:
-            logger.error(message)
-
-
-__all__ = [
-    # Main class
-    "LoadBalancer",
-    # Utility classes
-    "NodePerformanceTracker",
-    # Note: Load balancing models are available via DTO imports:
-    # from unrealon_sdk.src.dto.load_balancing import ...
-]