fix(ops): alert on stale gitea actions jobs

scripts/ops/docker-stats-textfile-exporter.py

@@ -15,6 +15,7 @@ import os
 import re
 import subprocess
 import tempfile
+from datetime import datetime, timezone
 from pathlib import Path
 
 
@@ -67,6 +68,17 @@ def _memory_bytes(value: str) -> float:
     return float(number) * scale
 
 
+def _started_at_seconds(value: str) -> float:
+    if not value:
+        return 0.0
+    try:
+        normalized = re.sub(r"\.(\d{6})\d+(Z|[+-]\d\d:\d\d)$", r".\1\2", value)
+        normalized = normalized.replace("Z", "+00:00")
+        return datetime.fromisoformat(normalized).astimezone(timezone.utc).timestamp()
+    except ValueError:
+        return 0.0
+
+
 def collect() -> str:
     stats = _run_json_lines([
         "docker",
@@ -102,6 +114,8 @@ def collect() -> str:
         "# TYPE docker_container_inspect_restart_count gauge",
         "# HELP docker_container_info Docker container inventory exposed by the textfile exporter.",
         "# TYPE docker_container_info gauge",
+        "# HELP docker_container_started_seconds Docker container start timestamp from Docker inspect.",
+        "# TYPE docker_container_started_seconds gauge",
     ]
 
     for row in stats:
@@ -114,6 +128,7 @@ def collect() -> str:
         nano_cpus = float(host_config.get("NanoCpus") or 0)
         memory_limit = float(host_config.get("Memory") or 0)
         restart_count = int(inspected.get("RestartCount") or 0)
+        started_seconds = _started_at_seconds(state.get("StartedAt", ""))
         labels = f'host="{_escape_label(HOST_LABEL)}",container_name="{_escape_label(name)}"'
         mem_current = (row.get("MemUsage") or "0 B / 0 B").split("/", 1)[0].strip()
         pids = row.get("PIDs") or "0"
@@ -123,6 +138,7 @@ def collect() -> str:
         lines.append(f"docker_container_memory_limit_bytes{{{labels}}} {memory_limit:.0f}")
         lines.append(f"docker_container_pids{{{labels}}} {int(float(pids))}")
         lines.append(f"docker_container_inspect_restart_count{{{labels}}} {restart_count}")
+        lines.append(f"docker_container_started_seconds{{{labels}}} {started_seconds:.6f}")
         lines.append(
             f'docker_container_info{{{labels},status="{_escape_label(state.get("Status", ""))}"}} 1'
         )