9db7fd917e
P6 硬化（5 项全部完成）:
- CDC 多实例水平扩展: _INSTANCE_ID + get_lag() 真实 lag 计算
- ExamCache Redis 化: key data_ana:exam:{exam_id}, TTL 30 天 + 内存 LRU fallback
- ClickHouse TTL 归档: 5 表均加 TTL（1-3 年），分区级删除
- Prometheus 监控: 18 个指标（CDC/CH/ExamCache/DataScope/gRPC/业务）
- readyz 深度硬化: 4 依赖超时检查（CH 1s/Redis 200ms/iam 2s/CDC lag<1000）
v2 新增 6 个 RPC（analytics.proto 扩展为 18 RPC）:
- GetStudentGrowth / GetAssignmentAnalysis / GetMasterySummary
- ListDiagnosticReports（占位，待 ai 服务）/ ListErrorBookItems / GetErrorBookStats
监控与可观测性: lifespan 预热 + gRPC ServerInterceptor + CDC 消费者指标
Docker 本地测试 19 项全部通过（healthz/readyz/metrics + 11 HTTP + 10 gRPC + ruff）
nextstep-v2.md: 上游需求对齐 + 下游要求（iam/core-edu/content/ai/SRE）
104 lines
3.3 KiB
Python
104 lines
3.3 KiB
Python
"""配置管理（pydantic-settings 完整配置项）.
|
||
|
||
对齐 02-architecture-design.md §15 配置清单.
|
||
环境变量前缀 DATA_ANA_（如 DATA_ANA_HTTP_PORT=3006）.
|
||
"""
|
||
|
||
from pydantic_settings import BaseSettings, SettingsConfigDict
|
||
|
||
|
||
class Settings(BaseSettings):
|
||
"""应用配置.
|
||
|
||
所有外部依赖（ClickHouse / Kafka / Redis / iam gRPC）均为可选：
|
||
未配置或不可达时服务进入降级模式，返回骨架数据 + details.degraded: true.
|
||
"""
|
||
|
||
# 服务
|
||
service_name: str = "data-ana"
|
||
http_port: int = 3006
|
||
grpc_port: int = 50055
|
||
log_level: str = "INFO"
|
||
dev_mode: bool = False
|
||
|
||
# ClickHouse（可选：留空则降级模式）
|
||
clickhouse_host: str = ""
|
||
clickhouse_port: int = 8123
|
||
clickhouse_database: str = "edu_analytics"
|
||
clickhouse_user: str = ""
|
||
clickhouse_password: str = ""
|
||
clickhouse_connect_timeout_ms: int = 3000
|
||
clickhouse_query_timeout_s: int = 3 # P4 退出标准 5s，查询 3s 超时降级
|
||
|
||
# Kafka
|
||
kafka_brokers: str = "" # 留空则不启动 CDC 消费者
|
||
kafka_consumer_group: str = "data-ana-cdc"
|
||
kafka_cdc_topics: str = (
|
||
"edu-cdc.next_edu_cloud.core_edu_grades,"
|
||
"edu-cdc.next_edu_cloud.core_edu_exams,"
|
||
"edu-cdc.next_edu_cloud.core_edu_homework_submissions,"
|
||
"edu-cdc.next_edu_cloud.core_edu_attendance,"
|
||
"edu-cdc.next_edu_cloud.classes,"
|
||
"edu-cdc.next_edu_cloud.iam_users,"
|
||
"edu-cdc.next_edu_cloud.content_knowledge_points"
|
||
)
|
||
kafka_mastery_topic: str = "edu.insight.mastery.updated"
|
||
kafka_warning_topic: str = "edu.insight.mastery.updated" # 复用 mastery topic（总裁裁决 §2.11）
|
||
kafka_ai_usage_topic: str = "edu.insight.ai.usage"
|
||
kafka_enable_auto_commit: bool = False # v2: 手动 commit（at-least-once）
|
||
kafka_auto_offset_reset: str = "latest"
|
||
kafka_producer_transactional_id: str = "data-ana-producer"
|
||
|
||
# iam gRPC
|
||
iam_grpc_endpoint: str = "" # 留空则使用降级兜底（按 role 映射 DataScope）
|
||
iam_grpc_timeout_s: int = 2
|
||
datascope_cache_ttl_s: int = 300 # 5min
|
||
|
||
# Redis
|
||
redis_url: str = "" # 留空则跳过缓存
|
||
redis_pool_size: int = 10
|
||
redis_socket_timeout_ms: int = 200
|
||
|
||
# OTel
|
||
otel_endpoint: str = "http://localhost:4318"
|
||
otel_service_name: str = "data-ana"
|
||
|
||
# 掌握度算法
|
||
mastery_window_size: int = 5
|
||
mastery_decay_base: float = 0.6
|
||
mastery_forgetting_half_life_days: int = 30
|
||
mastery_min_samples: int = 3
|
||
|
||
# 预警阈值
|
||
warning_low_mastery_threshold: float = 0.4
|
||
warning_critical_mastery_threshold: float = 0.2
|
||
warning_score_drop_percent: float = 0.2
|
||
warning_absent_per_week: int = 3
|
||
|
||
# 降级
|
||
degraded_mode_enabled: bool = True
|
||
|
||
# P6 readyz 深度硬化
|
||
readyz_clickhouse_timeout_s: float = 1.0
|
||
readyz_redis_timeout_s: float = 0.2
|
||
readyz_iam_grpc_timeout_s: float = 2.0
|
||
readyz_cdc_lag_threshold: int = 1000 # lag 超过此值判定 not_ready
|
||
|
||
# 向后兼容：旧代码引用 settings.port / settings.kafka_group_id
|
||
@property
|
||
def port(self) -> int:
|
||
return self.http_port
|
||
|
||
@property
|
||
def kafka_group_id(self) -> str:
|
||
return self.kafka_consumer_group
|
||
|
||
model_config = SettingsConfigDict(
|
||
env_file=".env",
|
||
env_prefix="DATA_ANA_",
|
||
extra="ignore",
|
||
)
|
||
|
||
|
||
settings = Settings()
|