[openclaw] Uncaught exception: Error: Unable to update lock within the stale threshold

[zhangqinghua2015]

Plugin Version

v1.1.0-beta.10

OpenClaw Version

2026.03.28

Bug Description

2026-03-30T15:06:07.905+08:00 [openclaw] Uncaught exception: Error: Unable to update lock within the stale threshold
    at /root/.openclaw/plugins/memory-lancedb-pro/node_modules/proper-lockfile/lib/lockfile.js:136:25
    at callback (/root/.openclaw/plugins/memory-lancedb-pro/node_modules/graceful-fs/polyfills.js:306:20)
    at FSReqCallback.oncomplete (node:fs:197:5)
    at FSReqCallback.callbackTrampoline (node:internal/async_hooks:130:17)

这个错误会导致openclaw gateway进程退出

Expected Behavior

配置如下：

"memory-lancedb-pro": {
        "enabled": true,
        "config": {
          "embedding": {
            "apiKey": "${JINA_API_KEY}",
            "model": "jina-embeddings-v5-text-small",
            "baseURL": "https://api.jina.ai/v1",
            "dimensions": 1024,
            "taskQuery": "retrieval.query",
            "taskPassage": "retrieval.passage",
            "normalized": true,
            "chunking": true
          },
          "autoCapture": true,
          "autoRecall": true,
          "smartExtraction": true,
          "llm": {
            "apiKey": "***",
            "model": "openai/gpt-oss-120b",
            "baseURL": "***",
            "auth": "api-key",
            "timeoutMs": 30000
          },
          "extractMinMessages": 2,
          "extractMaxChars": 8000,
          "sessionMemory": {
            "enabled": false,
            "messageCount": 15
          },
          "retrieval": {
            "rerank": "cross-encoder",
            "rerankApiKey": "${JINA_API_KEY}",
            "rerankModel": "jina-reranker-v3",
            "rerankEndpoint": "https://api.jina.ai/v1/rerank",
            "rerankProvider": "jina",
            "hardMinScore": 0.35,
            "mode": "hybrid",
            "vectorWeight": 0.7,
            "bm25Weight": 0.3,
            "minScore": 0.3,
            "candidatePoolSize": 20,
            "recencyHalfLifeDays": 14,
            "recencyWeight": 0.1,
            "filterNoise": true,
            "lengthNormAnchor": 500,
            "timeDecayHalfLifeDays": 60,
            "reinforcementFactor": 0.5,
            "maxHalfLifeMultiplier": 3
          },
          "mdMirror": {
            "enabled": true,
            "dir": "memory-md"
          },
          "enableManagementTools": false,
          "sessionStrategy": "memoryReflection",
          "autoRecallMinLength": 15,
          "autoRecallMinRepeated": 8,
          "autoRecallMaxItems": 3,
          "autoRecallMaxChars": 600,
          "autoRecallPerItemMaxChars": 180,
          "recallMode": "full"
        }
      },

Steps to Reproduce

和多agent每个agent有各自的workspace有关系吗？

Error Logs / Screenshots

Embedding Provider

None

OS / Platform

No response

[AliceLJY]

This is a proper-lockfile concurrency issue — it happens when multiple processes try to access the LanceDB database simultaneously (e.g., multiple agents or a gateway restart while a background task is running).

A few things to check:

1. Are you running multiple OpenClaw instances pointing to the same LanceDB path?
2. Did you recently restart the gateway while a memory operation was in progress?
3. Check if stale .lock files exist in your LanceDB directory: ls ~/.openclaw/memory/lancedb-pro/*.lock

As a workaround, you can stop OpenClaw, remove any stale lock files, and restart.

[zhangqinghua2015]

这是一个 proper-lockfile 并发问题——当多个进程同时尝试访问 LanceDB 数据库时就会发生这种情况（例如，多个代理或在后台任务运行时重启网关）。

需要检查以下几点：

1. 您是否运行了多个指向同一 LanceDB 路径的 OpenClaw 实例？
2. 您最近是否在内存操作进行期间重启过网关？
3. 检查 LanceDB 目录中是否存在过期的 .lock 文件： ls ~/.openclaw/memory/lancedb-pro/*.lock

作为一种变通方法，您可以停止 OpenClaw，删除任何过时的锁定文件，然后重新启动。

1、只有一个Openclaw实例，但是有多个agent，每个agent有自己的工作空间，LanceDB路径没指定设置
2、有异常重启过几次，不确定最开始的重启是不是在刚好内存操作期间
3、现在只看到有一个 .memory-write.lock
4、我之前是设置的 "autoRecall": false，后面设置"autoRecall": true 后出现这个错误，不确定有没有关系；现在我又改回了 "autoRecall": false，目前没有再出现这个错

[AliceLJY]

Thanks for the follow-up details! Since you only have one OpenClaw instance, the concurrent lock contention is likely coming from multiple agents all defaulting to the same LanceDB path (~/.openclaw/memory/lancedb-pro) and firing simultaneous reads when autoRecall: true.

Two options:

1. Per-agent DB paths: set dbPath per agent workspace so they each have an isolated DB and don't compete
2. Keep autoRecall: false and use explicit memory_recall calls only when needed — this avoids the concurrent read pattern entirely

The one remaining .memory-write.lock file is fine to leave; it will auto-expire on the next gateway start.

[zhangqinghua2015]

要怎么为每个agent设置 dbPath 呢？文档看起来dbPath似乎没办法为每个agent单独配置

[AliceLJY]

Hi @zhangqinghua2015, great question — and sorry for the delayed response!

You're right that per-agent dbPath is not currently supported. It's a global plugin setting, so all agents sharing the same gateway end up pointing to the same LanceDB directory. This means concurrent access from multiple agents can trigger the stale lock issue you're seeing, especially with autoRecall: true which adds read contention during prompt builds.

We recognize this is a real gap and will track per-agent dbPath support as a follow-up enhancement. In the meantime, keeping autoRecall: false is the most reliable workaround since it avoids the concurrent read contention that triggers the lock crash.

Thanks for your patience and for surfacing this — the docs should definitely be clearer about this limitation! 🙏

[jlin53882]

PR 準備好了

PR #517 修復了這個問題，包含兩個核心變更：

1. stale: 10s → 60s + retries max wait: 3s → 151s
2. ECOMPROMISED graceful fallback：不等於立刻崩，先等3秒再試一次

請 maintainer review 🙏