i/Aether
1
0
Fork 0
mirror of https://github.com/fawney19/Aether.git synced 2026-09-02 17:30:23 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix: 治理 Prometheus 指标基数爆炸和内存缓存无界增长

Browse Source 
- 移除 token/latency Prometheus 指标的 model 标签，避免 provider x model 笛卡尔积
- HealthMonitor 滑动窗口从 DB JSON 迁移至进程内存，减少写放大
- ModelCostService 三层缓存增加 500 条上限，超限时清空
- StickyPriority 粘性缓存和健康状态字典增加容量淘汰
- AffinityManager 请求锁字典增加 500 条上限，淘汰空闲锁
- 配额刷新/探测查询使用 defer/load_only 避免加载大 JSON 列
- Alembic 迁移清理 DB 中遗留的 request_results_window 数据
- 同步更新测试适配 batch_get_cooldowns 返回值和批量删除异步化
This commit is contained in:
fawney19
2026-03-10 10:40:50 +08:00
parent c9f0685b40
commit 7c580e843f
20 changed files with 241 additions and 85 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
tests/services/test_pool_manager_trace.py
View File

@@ -40,7 +40,7 @@ async def test_trace_attached_to_first_candidate() -> None:
patch(
"src.services.provider.pool.redis_ops.batch_get_cooldowns",
new_callable=AsyncMock,
return_value={"key-1": (None, None), "key-2": (None, None)},
return_value={"key-1": None, "key-2": None},
),
patch(
"src.services.provider.pool.redis_ops.get_lru_scores",
@@ -72,7 +72,7 @@ async def test_pool_extra_data_on_selected_candidate() -> None:
patch(
"src.services.provider.pool.redis_ops.batch_get_cooldowns",
new_callable=AsyncMock,
return_value={"key-1": (None, None)},
return_value={"key-1": None},
),
patch(
"src.services.provider.pool.redis_ops.get_lru_scores",
@@ -103,8 +103,8 @@ async def test_pool_extra_data_on_skipped_candidate() -> None:
"src.services.provider.pool.redis_ops.batch_get_cooldowns",
new_callable=AsyncMock,
return_value={
"key-1": ("rate_limited_429", 120),
"key-2": (None, None),
"key-1": "rate_limited_429",
"key-2": None,
},
),
patch(
@@ -122,7 +122,8 @@ async def test_pool_extra_data_on_skipped_candidate() -> None:
assert extra is not None
assert extra["pool_skip"]["type"] == "cooldown"
assert extra["pool_skip"]["cooldown_reason"] == "rate_limited_429"
assert extra["pool_skip"]["cooldown_ttl"] == 120
# TTL is not fetched on the scheduling hot path (include_ttl=False)
assert "cooldown_ttl" not in extra["pool_skip"]
@pytest.mark.asyncio
@@ -142,9 +143,9 @@ async def test_trace_build_summary_matches() -> None:
"src.services.provider.pool.redis_ops.batch_get_cooldowns",
new_callable=AsyncMock,
return_value={
"key-1": ("overloaded_529", 60),
"key-2": (None, None),
"key-3": (None, None),
"key-1": "overloaded_529",
"key-2": None,
"key-3": None,
},
),
patch(
@@ -181,7 +182,7 @@ async def test_trace_build_summary_uses_attempted_key_ids_when_provided() -> Non
patch(
"src.services.provider.pool.redis_ops.batch_get_cooldowns",
new_callable=AsyncMock,
return_value={"key-1": (None, None), "key-2": (None, None)},
return_value={"key-1": None, "key-2": None},
),
patch(
"src.services.provider.pool.redis_ops.get_lru_scores",
@@ -218,7 +219,7 @@ async def test_sticky_trace_info() -> None:
patch(
"src.services.provider.pool.redis_ops.batch_get_cooldowns",
new_callable=AsyncMock,
return_value={"key-1": (None, None), "key-2": (None, None)},
return_value={"key-1": None, "key-2": None},
),
patch(
"src.services.provider.pool.redis_ops.get_lru_scores",