feat: add agent run heartbeat to slo reports

This commit is contained in:
Your Name
2026-07-02 12:41:04 +08:00
parent 054379ac98
commit f9a385f333
5 changed files with 158 additions and 6 deletions
+1
View File
@@ -35,6 +35,7 @@ H5_PUBLIC_BASE_URL=http://127.0.0.1:5173
# MEMIND_AGENT_RUN_WORKER_POLL_MS=1000
# MEMIND_AGENT_RUN_WORKER_BATCH_SIZE=1
# MEMIND_AGENT_RUN_HEARTBEAT_MS=30000
# MEMIND_AGENT_RUN_HEARTBEAT_STALE_MS=90000
# MEMIND_AGENT_RUN_WORKER_EXPECT_RUNNING=0
# Runtime SLO 日报(默认每日 23:55 写 reports/runtime-slo)。
+15
View File
@@ -203,6 +203,21 @@ oldestRunningHeartbeatAgeMs=0 when no running rows
runningWithoutHeartbeatCount=0 during normal operation
```
Runtime SLO report should also expose the same heartbeat summary:
```bash
node /Users/john/Project/Memind/scripts/runtime-slo-report.mjs
```
Expected SLO fields:
```text
runtime.toolQueueSlo.heartbeatMs=30000
runtime.toolQueueSlo.oldestRunningHeartbeatAgeMs=0 when no running rows
runtime.toolQueueSlo.runningWithoutHeartbeatCount=0 during normal operation
failures=[] unless heartbeat age exceeds MEMIND_AGENT_RUN_HEARTBEAT_STALE_MS
```
Apply only when the listed running rows are known stale:
```bash
@@ -19,12 +19,12 @@
- Stream Controller 已具备 SSE headers、abort propagation、backpressure pipeline。
- Redis Router 已启用,承担 worker runtime state。
- Goose Worker Pool 已从固定单点走向四 worker 可观测调度。
- Aider/OpenHands 已从普通聊天默认能力中剥离,进入 code mode 和后端灰度门禁;P6.3 已把 code run 从 goosed session extension 外移到 `agent-run-v1` Tool Gateway 协议,P6.4 已完成 Aider 真实执行 canaryP6.5 已完成 OpenHands 真实执行 canaryP6.6 已完成 external worker 精确接管 code-run canaryP6.7 已加入 Tool Gateway 产物校验与输出审计,P6.8 已安装 external worker LaunchAgentP6.9 已完成带 validation 的 external worker 灰度 canaryP6.10 已加入 external worker 只读观测脚本,P6.11 已加入放量策略门禁,P6.12 已开启全用户长期灰度并通过普通测试用户真实路径,P6.13 已安装自动暂停 guardP6.15 已让 H5 code-run 自动补 receipt validation 并恢复后端 required validationP6.17/P6.18 已把 external worker 并发 2 通过 canary 并固化为当前 all-user gray 策略,P6.19 已加入 task-level artifact validationP8.1 已加入 stale running recoveryP8.2 已完成 Portal DB/Auth 瞬时错误兜底,P8.3 已加入 worker lease heartbeat。
- Aider/OpenHands 已从普通聊天默认能力中剥离,进入 code mode 和后端灰度门禁;P6.3 已把 code run 从 goosed session extension 外移到 `agent-run-v1` Tool Gateway 协议,P6.4 已完成 Aider 真实执行 canaryP6.5 已完成 OpenHands 真实执行 canaryP6.6 已完成 external worker 精确接管 code-run canaryP6.7 已加入 Tool Gateway 产物校验与输出审计,P6.8 已安装 external worker LaunchAgentP6.9 已完成带 validation 的 external worker 灰度 canaryP6.10 已加入 external worker 只读观测脚本,P6.11 已加入放量策略门禁,P6.12 已开启全用户长期灰度并通过普通测试用户真实路径,P6.13 已安装自动暂停 guardP6.15 已让 H5 code-run 自动补 receipt validation 并恢复后端 required validationP6.17/P6.18 已把 external worker 并发 2 通过 canary 并固化为当前 all-user gray 策略,P6.19 已加入 task-level artifact validationP8.1 已加入 stale running recoveryP8.2 已完成 Portal DB/Auth 瞬时错误兜底,P8.3 已加入 worker lease heartbeatP8.4 已把 heartbeat 纳入 runtime SLO 报告
- PG 和 MindSpace 仍保持生产数据边界,SLO 报告只做统计读取;P6.3-P6.15 不新增 schema migration,不删除或修改既有用户数据。
整体执行评分: 9.95 / 10。
可以支撑当前 H5 streaming 稳定性改造的基础目标。自动采样、worker sidecar heartbeat、SLO 只读快照、SLO 日报定时器、SLO 日报保留策略、first-token EWMA、first-token p50/p95 窗口趋势、Tool Gateway Queue v0、外部 worker 接管入口、真实 worker canary、后端 code-mode canary、code-run 用户级灰度 gate、H5 页面编辑 UI canary、P6.3 Tool Gateway 协议化、P6.4 Aider 真实 canary、P6.5 OpenHands 真实 canary、P6.6 external worker code-run canary、P6.7 Tool Gateway guardrails、P6.8 worker LaunchAgent、P6.9 validated external worker canary、P6.10 worker observability、P6.11 rollout policy gates、P6.12 all-user gray、P6.13 auto-pause guard、P6.15 H5 validation metadata、P6.17/P6.18 worker concurrency 2、P6.19 task-level artifact validation、P8.1 stale running recovery、P8.2 Portal DB/Auth transient hardeningP8.3 worker lease heartbeat 已经落地。P6.22 用户可见进度/失败说明按用户决策暂不开发,后续剩余差距转为更长时间 soak、自动报表告警和更细任务类型的校验策略扩展。
可以支撑当前 H5 streaming 稳定性改造的基础目标。自动采样、worker sidecar heartbeat、SLO 只读快照、SLO 日报定时器、SLO 日报保留策略、first-token EWMA、first-token p50/p95 窗口趋势、Tool Gateway Queue v0、外部 worker 接管入口、真实 worker canary、后端 code-mode canary、code-run 用户级灰度 gate、H5 页面编辑 UI canary、P6.3 Tool Gateway 协议化、P6.4 Aider 真实 canary、P6.5 OpenHands 真实 canary、P6.6 external worker code-run canary、P6.7 Tool Gateway guardrails、P6.8 worker LaunchAgent、P6.9 validated external worker canary、P6.10 worker observability、P6.11 rollout policy gates、P6.12 all-user gray、P6.13 auto-pause guard、P6.15 H5 validation metadata、P6.17/P6.18 worker concurrency 2、P6.19 task-level artifact validation、P8.1 stale running recovery、P8.2 Portal DB/Auth transient hardeningP8.3 worker lease heartbeat 和 P8.4 heartbeat SLO 已经落地。P6.22 用户可见进度/失败说明按用户决策暂不开发,后续剩余差距转为 24h soak 观察和更细任务类型的校验策略扩展。
## 实测结果
@@ -1016,6 +1016,42 @@ Data boundary:
- worker check `ok=true`, `expected=running`
- guard `shouldPause=false`
### P8.4 Long-running Gray Soak and Heartbeat SLO
结果: 通过,runtime SLO JSON/Markdown 已加入 agent-run heartbeat 摘要和 failure 条件,生产只读与临时报表写入验证均通过。
- `runtime.toolQueueSlo` 新增:
- `heartbeatMs`
- `maxConcurrentRuns`
- `statusCounts`
- `oldestRunningStartedAt`
- `oldestRunningAgeMs`
- `oldestRunningHeartbeatAt`
- `oldestRunningHeartbeatAgeMs`
- `runningWithoutHeartbeatCount`
- `latestRunningRun`
- failure 条件新增:
- `agent_run_heartbeat_stale_<age>`
- `agent_run_missing_heartbeat_<count>`
- Markdown Tool Queue 段落新增 heartbeat summary。
- 测试:
- `node --check scripts/runtime-slo-report.mjs`
- 源码目录只读 SLO `ok=true`, `failures=[]`
- 生产部署:
- backup `/Users/john/Project/memind_backups/20260702-123947-p84-heartbeat-slo`
- deployed `scripts/runtime-slo-report.mjs`
- no service restart
- 生产验证:
- production read-only SLO `ok=true`, `failures=[]`
- `toolQueueSlo.heartbeatMs=30000`
- `toolQueueSlo.statusCounts={}`
- `toolQueueSlo.runningWithoutHeartbeatCount=0`
- temporary JSON/Markdown report files include heartbeat summary fields
- 定时器:
- `cn.tkmind.memind-runtime-slo-report`
- remains scheduled at `23:55`
- writes with `--write-report --prune --retention-days 30`
## 下一步执行建议
### P6.22 Task Artifact UX and Failure Messages
@@ -1026,6 +1062,5 @@ Data boundary:
### P8.4 Long-running Gray Soak and Heartbeat SLO
- 基于 heartbeat 字段做一轮 24h all-user gray soak
- SLO 日报增加 running heartbeat age 和 missing heartbeat count
- guard 阈值保守观察,不扩大并发。
- 第一阶段 SLO 字段已完成
- 下一步是 24h all-user gray soak 观察,不扩大并发
@@ -30,6 +30,7 @@
- P8.1 Queue Lease / Stuck Run Recovery: 已完成第一步,worker dispatch 前自动回收超时 running run,新增 `--recover-stale` dry-run/apply 运维入口,生产 synthetic stale run 验证通过。
- P8.2 Portal DB/Auth Transient Error Hardening: 已完成第一步,session attach、`/auth/status` 和 API auth middleware 已捕获 DB/auth 瞬时错误,生产部署后 live health、auth/status、runtime/status、guard 和 SLO 验证通过。
- P8.3 Worker Lease Heartbeat: 已完成第一步,running run 执行期间写入 `worker_heartbeat` eventruntime/status、worker check、guard 和 stale recovery 均优先使用 heartbeat age,生产 synthetic 与真实 worker canary 验证通过。
- P8.4 Long-running Gray Soak and Heartbeat SLO: 已完成第一步,runtime SLO JSON/Markdown 报告加入 agent-run heartbeat 摘要和 failure 条件,生产只读与临时报表写入验证通过;继续保持并发 2 soak,不扩大灰度。
- P5.15 Active Stream TTL Reconcile: 已按用户要求跳过,暂不做报表/定时 reconcile。
- P5 Worker Pool 运维化: 已完成第一步,Redis Router 支持 worker drain。
- P5.9 First-token Latency EWMA: 已完成,StreamController 会把首个 SSE chunk 延迟写入 RedisSLO 报告已展示。
@@ -3015,6 +3016,73 @@ runtime/status:
- heartbeat data includes `pid=31341`, `attempt=1`, `heartbeatMs=30000`
- `tool_gateway_validation` passed for receipt and `public/p83-heartbeat-real-20260702043642.html`
### 2026-07-02 P8.4 Long-running Gray Soak and Heartbeat SLO
目标:
- 不扩大并发、不做 UI,先把 P8.3 heartbeat 纳入每日 SLO 报告。
- 让 23:55 定时日报和手动 `runtime-slo-report.mjs` 都能暴露 running heartbeat 风险。
- 为后续 24h all-user gray soak 提供稳定观测字段。
改动:
- `scripts/runtime-slo-report.mjs`:
- `runtime.toolQueueSlo` 新增 heartbeat 摘要:
- `heartbeatMs`
- `maxConcurrentRuns`
- `statusCounts`
- `oldestRunningStartedAt`
- `oldestRunningAgeMs`
- `oldestRunningHeartbeatAt`
- `oldestRunningHeartbeatAgeMs`
- `runningWithoutHeartbeatCount`
- `latestRunningRun`
- failure 条件新增:
- `agent_run_heartbeat_stale_<age>`: running heartbeat age 超过阈值。
- `agent_run_missing_heartbeat_<count>`: running run 长时间缺失 heartbeat。
- Markdown 报告的 Tool Queue 段落展示:
- `heartbeatMs`
- `oldestRunningHeartbeatAgeMs`
- `runningWithoutHeartbeatCount`
- `.env.example`:
- 记录 `MEMIND_AGENT_RUN_HEARTBEAT_STALE_MS=90000`
- `docs/agent-run-worker-rollout-runbook.md`:
- 记录 heartbeat SLO 检查命令和期望字段。
测试:
- `node --check scripts/runtime-slo-report.mjs` 通过。
- 源码目录只读 SLO:
- `ok=true`
- `failures=[]`
- `toolQueueSlo.heartbeatMs=30000`
- `toolQueueSlo.runningWithoutHeartbeatCount=0`
生产部署:
- 备份:
- `/Users/john/Project/memind_backups/20260702-123947-p84-heartbeat-slo`
- 已部署:
- `scripts/runtime-slo-report.mjs`
- 未重启 Portal、worker、goosed 或 H5。
- 现有 SLO LaunchAgent 保持:
- `cn.tkmind.memind-runtime-slo-report`
- 每日 `23:55`
- `--write-report --prune --retention-days 30`
生产验证:
- 生产只读 SLO:
- `ok=true`
- `failures=[]`
- `toolQueueSlo.heartbeatMs=30000`
- `toolQueueSlo.statusCounts={}`
- `toolQueueSlo.oldestRunningHeartbeatAgeMs=0`
- `toolQueueSlo.runningWithoutHeartbeatCount=0`
- 临时 report-dir 写入验证:
- JSON 和 Markdown 均生成成功。
- 两种格式均包含 `heartbeatMs``oldestRunningHeartbeatAgeMs``runningWithoutHeartbeatCount`
## 回滚策略
- P0: 修改前保留 `server.mjs` 备份;如启动失败,恢复备份并 `launchctl kickstart` Portal。
+34 -1
View File
@@ -224,6 +224,7 @@ async function readRedisSummary(namespace, redisUrl) {
function summarizeRuntime(runtimeJson) {
const workers = runtimeJson?.router?.workers ?? [];
const toolQueue = runtimeJson?.toolRuntime?.queue ?? null;
const staleMetricMs = 2 * 60 * 1000;
const staleHeartbeatMs = Number(process.env.MEMIND_RUNTIME_HEARTBEAT_STALE_MS || 60 * 1000);
const now = Date.now();
@@ -231,7 +232,20 @@ function summarizeRuntime(runtimeJson) {
routerEnabled: Boolean(runtimeJson?.router?.enabled),
publicBaseUrl: runtimeJson?.publicBaseUrl ?? null,
toolRuntime: runtimeJson?.toolRuntime ?? null,
toolQueue: runtimeJson?.toolRuntime?.queue ?? null,
toolQueue,
toolQueueSlo: toolQueue
? {
heartbeatMs: Number(toolQueue.heartbeatMs ?? 0),
maxConcurrentRuns: Number(toolQueue.maxConcurrentRuns ?? 0),
statusCounts: toolQueue.statusCounts ?? {},
oldestRunningStartedAt: toolQueue.oldestRunningStartedAt ?? null,
oldestRunningAgeMs: Number(toolQueue.oldestRunningAgeMs ?? 0),
oldestRunningHeartbeatAt: toolQueue.oldestRunningHeartbeatAt ?? null,
oldestRunningHeartbeatAgeMs: Number(toolQueue.oldestRunningHeartbeatAgeMs ?? 0),
runningWithoutHeartbeatCount: Number(toolQueue.runningWithoutHeartbeatCount ?? 0),
latestRunningRun: toolQueue.latestRunningRun ?? null,
}
: null,
workers: workers.map((worker) => ({
id: worker.id,
healthy: runtimeJson?.targets?.find((target) => target.target === worker.target)?.healthy ?? null,
@@ -325,6 +339,21 @@ if (runtimeSummary?.toolQueue?.error) failures.push('tool_queue_status_error');
if (runtimeSummary?.toolQueue?.inFlight > runtimeSummary?.toolQueue?.maxConcurrentRuns) {
failures.push('tool_queue_concurrency_exceeded');
}
const agentRunHeartbeatStaleMs = Number(
process.env.MEMIND_AGENT_RUN_HEARTBEAT_STALE_MS ||
Math.max(90 * 1000, Number(runtimeSummary?.toolQueueSlo?.heartbeatMs ?? 0) * 3),
);
if (
runtimeSummary?.toolQueueSlo?.oldestRunningHeartbeatAgeMs > agentRunHeartbeatStaleMs
) {
failures.push(`agent_run_heartbeat_stale_${runtimeSummary.toolQueueSlo.oldestRunningHeartbeatAgeMs}`);
}
if (
Number(runtimeSummary?.toolQueueSlo?.runningWithoutHeartbeatCount ?? 0) > 0 &&
Number(runtimeSummary?.toolQueueSlo?.oldestRunningAgeMs ?? 0) > agentRunHeartbeatStaleMs
) {
failures.push(`agent_run_missing_heartbeat_${runtimeSummary.toolQueueSlo.runningWithoutHeartbeatCount}`);
}
const reportDir = path.resolve(args.reportDir || path.join(appRoot, 'reports', 'runtime-slo'));
if (args.writeReport || args.prune) {
@@ -392,6 +421,10 @@ function markdownReport(payload) {
'',
'## Tool Queue',
'',
`- heartbeatMs: ${payload.runtime?.toolQueueSlo?.heartbeatMs ?? 0}`,
`- oldestRunningHeartbeatAgeMs: ${payload.runtime?.toolQueueSlo?.oldestRunningHeartbeatAgeMs ?? 0}`,
`- runningWithoutHeartbeatCount: ${payload.runtime?.toolQueueSlo?.runningWithoutHeartbeatCount ?? 0}`,
'',
'```json',
JSON.stringify(payload.runtime?.toolQueue ?? null, null, 2),
'```',