From c206f13216b6d60dd5e2b52e5bdde4ce2aa491a9 Mon Sep 17 00:00:00 2001 From: Codex Date: Fri, 15 May 2026 00:18:42 +0000 Subject: [PATCH] fix: cap postgres connection pools --- docker-compose.yml | 5 +++++ docs/reference/deployment.md | 6 ++++++ src/components/backend-core/src/index.ts | 5 ++++- src/components/database/config/postgresql.conf | 6 +++--- src/components/microservices/baidu-netdisk/src/index.ts | 9 ++++++++- .../microservices/code-queue/docker-compose.d601.yml | 1 + src/components/microservices/code-queue/src/index.ts | 4 +++- src/components/microservices/code-queue/src/types.ts | 1 + src/components/microservices/oa-event-flow/src/index.ts | 9 ++++++++- .../microservices/project-manager/src/index.ts | 9 ++++++++- 10 files changed, 47 insertions(+), 8 deletions(-) diff --git a/docker-compose.yml b/docker-compose.yml index eae3a63f..ba0d12fc 100644 --- a/docker-compose.yml +++ b/docker-compose.yml @@ -56,6 +56,7 @@ services: PROVIDER_PORT: "8081" DATABASE_URL: "postgres://${UNIDESK_DATABASE_USER}:${UNIDESK_DATABASE_PASSWORD}@database:5432/${UNIDESK_DATABASE_NAME}" PROVIDER_TOKEN: "${UNIDESK_PROVIDER_TOKEN}" + DATABASE_POOL_MAX: "${UNIDESK_BACKEND_CORE_DATABASE_POOL_MAX:-4}" HEARTBEAT_TIMEOUT_MS: "${UNIDESK_HEARTBEAT_TIMEOUT_MS}" TASK_PENDING_TIMEOUT_MS: "${UNIDESK_TASK_PENDING_TIMEOUT_MS:-600000}" DATABASE_VOLUME_NAME: "${UNIDESK_DATABASE_VOLUME}" @@ -89,6 +90,7 @@ services: PORT: "4211" TODO_NOTE_BACKEND_ONLY: "1" DATABASE_URL: "postgres://${UNIDESK_DATABASE_USER}:${UNIDESK_DATABASE_PASSWORD}@database:5432/${UNIDESK_DATABASE_NAME}" + PGAPPNAME: "unidesk-todo-note" TODO_NOTE_LOGS_DIR: "logs" LOG_FILE: "/var/log/unidesk/${UNIDESK_LOG_DAY}/${UNIDESK_LOG_PREFIX}_todo-note.jsonl" TODO_NOTE_LOG_PATH: "/var/log/unidesk/${UNIDESK_LOG_DAY}/${UNIDESK_LOG_PREFIX}_todo-note.jsonl" @@ -127,6 +129,7 @@ services: HOST: "0.0.0.0" PORT: "4255" DATABASE_URL: "postgres://${UNIDESK_DATABASE_USER}:${UNIDESK_DATABASE_PASSWORD}@database:5432/${UNIDESK_DATABASE_NAME}" + DATABASE_POOL_MAX: "${UNIDESK_OA_EVENT_FLOW_DATABASE_POOL_MAX:-2}" PIPELINE_OA_BRIDGE_BASE_URL: "http://backend-core:8080/api/microservices/pipeline/proxy" PIPELINE_OA_BRIDGE_INTERVAL_MS: "${UNIDESK_PIPELINE_OA_BRIDGE_INTERVAL_MS:-15000}" PIPELINE_OA_BRIDGE_RUN_LIMIT: "${UNIDESK_PIPELINE_OA_BRIDGE_RUN_LIMIT:-50}" @@ -155,6 +158,7 @@ services: HOST: "0.0.0.0" PORT: "4233" DATABASE_URL: "postgres://${UNIDESK_DATABASE_USER}:${UNIDESK_DATABASE_PASSWORD}@database:5432/${UNIDESK_DATABASE_NAME}" + DATABASE_POOL_MAX: "${UNIDESK_PROJECT_MANAGER_DATABASE_POOL_MAX:-1}" LOG_FILE: "/var/log/unidesk/${UNIDESK_LOG_DAY}/${UNIDESK_LOG_PREFIX}_project-manager.jsonl" UNIDESK_LOG_RETENTION_BYTES: "${UNIDESK_LOG_RETENTION_BYTES:-1GiB}" volumes: @@ -180,6 +184,7 @@ services: HOST: "0.0.0.0" PORT: "4244" DATABASE_URL: "postgres://${UNIDESK_DATABASE_USER}:${UNIDESK_DATABASE_PASSWORD}@database:5432/${UNIDESK_DATABASE_NAME}" + DATABASE_POOL_MAX: "${UNIDESK_BAIDU_NETDISK_DATABASE_POOL_MAX:-2}" BAIDU_NETDISK_CLIENT_ID: "${UNIDESK_BAIDU_NETDISK_CLIENT_ID:-}" BAIDU_NETDISK_CLIENT_SECRET: "${UNIDESK_BAIDU_NETDISK_CLIENT_SECRET:-}" BAIDU_NETDISK_TOKEN_KEY: "${UNIDESK_BAIDU_NETDISK_TOKEN_KEY:-}" diff --git a/docs/reference/deployment.md b/docs/reference/deployment.md index 18f5af7a..d6313390 100644 --- a/docs/reference/deployment.md +++ b/docs/reference/deployment.md @@ -55,6 +55,12 @@ frontend 的 Docker 上线顺序为:先运行必要的本地校验,例如 `b Code Queue 已从主 server 迁移到 D601,但仍必须保持明确的 memory/swap 硬上限,默认 `CODE_QUEUE_MAX_ACTIVE_QUEUES=0` 以恢复 queue 间并行,仍保持 `CODE_QUEUE_IN_MEMORY_OUTPUT_RECORDS=10`、`CODE_QUEUE_IN_MEMORY_EVENT_RECORDS=10` 这类小热窗口;任务历史、队列统计、Trace/output 读取和 `/health` 摘要必须优先从 PostgreSQL 直读或聚合,不能为了性能便利在 Bun 进程内缓存全量历史。任何提高 Code Queue 热窗口、日志缓冲、Playwright/Codex 子进程常驻规模或容器上限的变更,或把 `CODE_QUEUE_MAX_ACTIVE_QUEUES` 显式改成正数,都必须在同一任务里说明 D601 资源预算来源,并通过 D601 `docker inspect code-queue-backend`、`docker stats --no-stream code-queue-backend`、`microservice health code-queue` 和对应 E2E 证明未重新引入内存爆炸风险。 +## Database Connection Budget + +主 PostgreSQL 的内存预算按“少量长驻服务连接池 + 短查询按需连接”设计,不允许每个 Bun 服务沿用默认 8 到 10 个连接。`backend-core` 默认 `DATABASE_POOL_MAX=4`,主 server 上的 `oa-event-flow`、`baidu-netdisk` 默认 `DATABASE_POOL_MAX=2`,`project-manager` 默认 `DATABASE_POOL_MAX=1`,D601 `code-queue` 默认 `CODE_QUEUE_DATABASE_POOL_MAX=2`;如需提高任一连接池上限,必须同时说明并发 SQL 需求、验证 `pg_stat_activity` 中该服务没有长期 idle 堆积,并确认 `max_connections` 仍有足够余量。PostgreSQL 基础配置固定保守值:`shared_buffers=128MB`、`work_mem=4MB`、`maintenance_work_mem=64MB`、`max_connections=50`,避免主 server 低内存环境被空闲 backend 和过大的 per-query 内存预算挤占。 + +排查 PostgreSQL 内存时以 `docker stats unidesk-database`、`pg_stat_activity` 分组和 `pg_settings` 为准;主机 `ps` 中每个 `postgres` 进程的 RSS 会重复计入共享内存,不能把所有 backend RSS 简单相加当作真实容器占用。所有 UniDesk PostgreSQL 客户端都必须设置可识别的 `application_name`,便于按服务统计连接数、状态和慢查询归属。 + ## Database Volume 架构要求数据库使用 10 GB named volume;当前实现将 volume 命名为 `unidesk_pgdata_10gb` 以固定生命周期。Docker named volume 默认不强制容量上限;如需硬配额,应在主机存储层或 Docker volume driver 层配置。CLI server 控制只能使用不删除 volume 的 `down` / `up` 流程,禁止使用 `down -v`、`docker volume rm` 或删除 `unidesk_pgdata_10gb`。 diff --git a/src/components/backend-core/src/index.ts b/src/components/backend-core/src/index.ts index 7f5ce456..e5b36a7a 100644 --- a/src/components/backend-core/src/index.ts +++ b/src/components/backend-core/src/index.ts @@ -41,6 +41,7 @@ interface RuntimeConfig { baiduNetdiskInternalUrl: string; microservices: MicroserviceConfig[]; logFile: string; + databasePoolMax: number; } interface MicroserviceConfig { @@ -204,9 +205,10 @@ const serviceStartedAt = new Date(); const config = readConfig(); const logger = createLogger("backend-core", config.logFile); const sql = postgres(config.databaseUrl, { - max: 8, + max: config.databasePoolMax, idle_timeout: 20, connect_timeout: 10, + connection: { application_name: "unidesk-backend-core" }, }); let dbReady = false; @@ -363,6 +365,7 @@ function readConfig(): RuntimeConfig { baiduNetdiskInternalUrl: process.env.BAIDU_NETDISK_INTERNAL_URL || "http://baidu-netdisk:4244", microservices: readMicroservicesEnv(), logFile: requiredEnv("LOG_FILE"), + databasePoolMax: Math.max(1, Math.min(16, readOptionalNumberEnv("DATABASE_POOL_MAX", 4))), }; } diff --git a/src/components/database/config/postgresql.conf b/src/components/database/config/postgresql.conf index cd3a7693..2d24566b 100644 --- a/src/components/database/config/postgresql.conf +++ b/src/components/database/config/postgresql.conf @@ -1,6 +1,6 @@ # UniDesk keeps PostgreSQL mostly stock; runtime logging options are injected by docker-compose.yml. -shared_buffers = '256MB' -work_mem = '16MB' +shared_buffers = '128MB' +work_mem = '4MB' maintenance_work_mem = '64MB' -max_connections = 100 +max_connections = 50 listen_addresses = '*' diff --git a/src/components/microservices/baidu-netdisk/src/index.ts b/src/components/microservices/baidu-netdisk/src/index.ts index fe377e86..32d9b94e 100644 --- a/src/components/microservices/baidu-netdisk/src/index.ts +++ b/src/components/microservices/baidu-netdisk/src/index.ts @@ -17,6 +17,7 @@ interface RuntimeConfig { port: number; databaseUrl: string; logFile: string; + databasePoolMax: number; clientId: string; clientSecret: string; tokenKey: string; @@ -113,6 +114,7 @@ function configFromEnv(): RuntimeConfig { port: Number(process.env.PORT || 4244), databaseUrl, logFile: process.env.LOG_FILE || "", + databasePoolMax: Math.max(1, Math.min(8, Number(process.env.DATABASE_POOL_MAX || 2) || 2)), clientId: process.env.BAIDU_NETDISK_CLIENT_ID || process.env.BAIDU_NETDISK_APP_KEY || "", clientSecret: process.env.BAIDU_NETDISK_CLIENT_SECRET || process.env.BAIDU_NETDISK_SECRET_KEY || "", tokenKey: process.env.BAIDU_NETDISK_TOKEN_KEY || "", @@ -124,7 +126,12 @@ function configFromEnv(): RuntimeConfig { } const config = configFromEnv(); -const sql = postgres(config.databaseUrl, { max: 8, idle_timeout: 20, connect_timeout: 10 }); +const sql = postgres(config.databaseUrl, { + max: config.databasePoolMax, + idle_timeout: 20, + connect_timeout: 10, + connection: { application_name: "unidesk-baidu-netdisk" }, +}); const logWriter = config.logFile ? createHourlyJsonlWriter({ baseLogFile: config.logFile, diff --git a/src/components/microservices/code-queue/docker-compose.d601.yml b/src/components/microservices/code-queue/docker-compose.d601.yml index eef7e294..25c831b8 100644 --- a/src/components/microservices/code-queue/docker-compose.d601.yml +++ b/src/components/microservices/code-queue/docker-compose.d601.yml @@ -28,6 +28,7 @@ services: CODE_QUEUE_APPROVAL_POLICY: "${CODE_QUEUE_APPROVAL_POLICY:-never}" CODE_QUEUE_MAX_ATTEMPTS: "${CODE_QUEUE_MAX_ATTEMPTS:-99}" CODE_QUEUE_MAX_ACTIVE_QUEUES: "${CODE_QUEUE_MAX_ACTIVE_QUEUES:-0}" + CODE_QUEUE_DATABASE_POOL_MAX: "${CODE_QUEUE_DATABASE_POOL_MAX:-2}" NODE_OPTIONS: "${CODE_QUEUE_NODE_OPTIONS:---max-old-space-size=1024}" CODE_QUEUE_IN_MEMORY_OUTPUT_RECORDS: "${CODE_QUEUE_IN_MEMORY_OUTPUT_RECORDS:-10}" CODE_QUEUE_IN_MEMORY_EVENT_RECORDS: "${CODE_QUEUE_IN_MEMORY_EVENT_RECORDS:-10}" diff --git a/src/components/microservices/code-queue/src/index.ts b/src/components/microservices/code-queue/src/index.ts index 2e32c941..7c56f3dc 100644 --- a/src/components/microservices/code-queue/src/index.ts +++ b/src/components/microservices/code-queue/src/index.ts @@ -184,9 +184,10 @@ let persistDirty = false; let shutdownRequested = false; let serviceReady = false; const sql: SqlClient = postgres(config.databaseUrl, { - max: 4, + max: config.databasePoolMax, idle_timeout: 20, connect_timeout: 10, + connection: { application_name: "unidesk-code-queue" }, }); let databaseReady = false; let databaseLastError: string | null = null; @@ -322,6 +323,7 @@ function readConfig(): RuntimeConfig { judgeMaxTokens: Math.max(800, Math.min(4000, envNumber("MINIMAX_JUDGE_MAX_TOKENS", 1800))), turnNoActivityTimeoutMs: Math.max(60_000, Math.min(30 * 60_000, envNumber("CODEX_TURN_NO_ACTIVITY_TIMEOUT_MS", 6 * 60_000))), databaseUrl: envRequiredString("DATABASE_URL"), + databasePoolMax: Math.max(1, Math.min(8, envNumber("CODE_QUEUE_DATABASE_POOL_MAX", envNumber("DATABASE_POOL_MAX", 2)))), databaseFlushIntervalMs: Math.max(100, Math.min(10_000, envNumber("CODE_QUEUE_DATABASE_FLUSH_INTERVAL_MS", 1000))), oaEventFlowBaseUrl: envString("OA_EVENT_FLOW_BASE_URL", "http://oa-event-flow:4255").replace(/\/+$/u, ""), notifyClaudeQqEnabled: envBool("CODE_QUEUE_NOTIFY_CLAUDEQQ_ENABLED", false), diff --git a/src/components/microservices/code-queue/src/types.ts b/src/components/microservices/code-queue/src/types.ts index 512f6350..2fa8d287 100644 --- a/src/components/microservices/code-queue/src/types.ts +++ b/src/components/microservices/code-queue/src/types.ts @@ -64,6 +64,7 @@ export interface RuntimeConfig { judgeMaxTokens: number; turnNoActivityTimeoutMs: number; databaseUrl: string; + databasePoolMax: number; databaseFlushIntervalMs: number; oaEventFlowBaseUrl: string; notifyClaudeQqEnabled: boolean; diff --git a/src/components/microservices/oa-event-flow/src/index.ts b/src/components/microservices/oa-event-flow/src/index.ts index 663a0694..12d347df 100644 --- a/src/components/microservices/oa-event-flow/src/index.ts +++ b/src/components/microservices/oa-event-flow/src/index.ts @@ -12,6 +12,7 @@ interface RuntimeConfig { port: number; databaseUrl: string; logFile: string; + databasePoolMax: number; pipelineBridgeBaseUrl: string; pipelineBridgeIntervalMs: number; pipelineBridgeRunLimit: number; @@ -179,6 +180,7 @@ function configFromEnv(): RuntimeConfig { port: envNumber("PORT", 4255), databaseUrl: envRequiredString("DATABASE_URL"), logFile: envString("LOG_FILE", "/var/log/unidesk/oa-event-flow.jsonl"), + databasePoolMax: Math.max(1, Math.min(8, envNumber("DATABASE_POOL_MAX", 2))), pipelineBridgeBaseUrl: envString("PIPELINE_OA_BRIDGE_BASE_URL", "").replace(/\/+$/u, ""), pipelineBridgeIntervalMs: envNumber("PIPELINE_OA_BRIDGE_INTERVAL_MS", 15_000), pipelineBridgeRunLimit: envNumber("PIPELINE_OA_BRIDGE_RUN_LIMIT", 50), @@ -186,7 +188,12 @@ function configFromEnv(): RuntimeConfig { } const config = configFromEnv(); -const sql: SqlClient = postgres(config.databaseUrl, { max: 8, idle_timeout: 20, connect_timeout: 10 }); +const sql: SqlClient = postgres(config.databaseUrl, { + max: config.databasePoolMax, + idle_timeout: 20, + connect_timeout: 10, + connection: { application_name: "unidesk-oa-event-flow" }, +}); const pipelineBridgeState: PipelineBridgeState = { enabled: config.pipelineBridgeBaseUrl.length > 0, mode: "snapshot", diff --git a/src/components/microservices/project-manager/src/index.ts b/src/components/microservices/project-manager/src/index.ts index f2a702c9..7b3b057c 100644 --- a/src/components/microservices/project-manager/src/index.ts +++ b/src/components/microservices/project-manager/src/index.ts @@ -13,6 +13,7 @@ interface RuntimeConfig { port: number; databaseUrl: string; logFile: string; + databasePoolMax: number; } interface ProjectRow { @@ -90,11 +91,17 @@ function configFromEnv(): RuntimeConfig { port: Number(process.env.PORT || 4233), databaseUrl, logFile: process.env.LOG_FILE || "", + databasePoolMax: Math.max(1, Math.min(8, Number(process.env.DATABASE_POOL_MAX || 1) || 1)), }; } const config = configFromEnv(); -const sql = postgres(config.databaseUrl, { max: 8, idle_timeout: 20, connect_timeout: 10 }); +const sql = postgres(config.databaseUrl, { + max: config.databasePoolMax, + idle_timeout: 20, + connect_timeout: 10, + connection: { application_name: "unidesk-project-manager" }, +}); const logWriter = config.logFile ? createHourlyJsonlWriter({ baseLogFile: config.logFile,