Dbochman/openclaw-cron-ghost-jobs

OpenClaw Cron Ghost Jobs

Problem

After removing a cron job from ~/.openclaw/cron/jobs.json, the job continues to execute on its previous schedule. This causes duplicate deliveries when a new replacement job is added alongside the still-running old one.

Context / Trigger Conditions

• A cron job was removed from jobs.json but still fires
• Duplicate messages delivered to the same target (e.g., two morning briefings)
• Old/legacy jobs run alongside their replacements after a migration
• ls ~/.openclaw/cron/runs/ shows JSONL files for job IDs not in jobs.json

Root Cause

OpenClaw's cron subsystem persists run state in per-job JSONL files at ~/.openclaw/cron/runs/<job-id>.jsonl. Each run appends a JSON line with nextRunAtMs indicating when to fire next. The cron scheduler reads these files independently — if a run file exists with a future nextRunAtMs, the job will execute even if its definition is gone from jobs.json.

Removing a job from jobs.json does NOT automatically clean up its run file.

Diagnosis

Step 1: Identify orphan run files

ssh dylans-mac-mini 'python3 -c "
import json, os
with open(\"/Users/dbochman/.openclaw/cron/jobs.json\") as f:
    job_ids = {j[\"id\"] for j in json.load(f)[\"jobs\"]}
for f in sorted(os.listdir(\"/Users/dbochman/.openclaw/cron/runs/\")):
    run_id = f.replace(\".jsonl\", \"\")
    if run_id not in job_ids:
        print(f\"ORPHAN: {f}\")
"'

Step 2: Check which orphans have future runs scheduled

ssh dylans-mac-mini 'python3 -c "
import json, os, time
runs_dir = \"/Users/dbochman/.openclaw/cron/runs/\"
now_ms = int(time.time() * 1000)
for f in sorted(os.listdir(runs_dir)):
    with open(os.path.join(runs_dir, f)) as fh:
        lines = fh.readlines()
    if lines:
        last = json.loads(lines[-1])
        next_run = last.get(\"nextRunAtMs\", 0)
        if next_run > now_ms:
            print(f\"ACTIVE GHOST: {f} nextRun={next_run}\")
"'

Solution

Delete the orphan run files:

ssh dylans-mac-mini 'trash ~/.openclaw/cron/runs/<orphan-job-id>.jsonl'

No gateway restart is needed — the cron subsystem will simply not find the run file on its next timer tick.

Prevention

When removing cron jobs from jobs.json, always also delete their run files:

# Remove from jobs.json (edit the file)
# Then clean up the run file
trash ~/.openclaw/cron/runs/<job-id>.jsonl

Verification

After cleanup:

# Only active job IDs should have run files
ls ~/.openclaw/cron/runs/
# Compare against:
python3 -c "import json; [print(j['id']) for j in json.load(open('/Users/dbochman/.openclaw/cron/jobs.json'))['jobs']]"

Notes

• Run files are append-only JSONL — each line is a run record with status, summary, usage, nextRunAtMs
• One-shot jobs (deleteAfterRun: true) clean up their own run files after execution
• Recurring cron jobs (kind: "cron") accumulate run history indefinitely
• The gateway hot-reloads jobs.json but does NOT scan for orphan run files