Add stats to enable monitoring for non-running containers

.gitignore

@@ -1 +1,3 @@
+.mypy_cache/
 .pyenv/
+.vscode/

dockstat.py

@@ -5,19 +5,21 @@ Module to act as a Prometheus Exporter for Docker containers with a
 """
 
 import argparse
+import os
 import os.path
 import sys
 from http.server import HTTPServer
 
-import docker
-from prometheus_client import (
+import docker  # type: ignore
+import numpy
+from prometheus_client import (  # type: ignore
     CollectorRegistry,
     Gauge,
-    generate_latest,
     MetricsHandler,
+    generate_latest,
 )
 
-LISTEN_PORT = 8080
+LISTEN_PORT = int(os.environ.get('LISTEN_PORT', 8080))
 HEALTHY_STR = 'healthy'
 
 
@@ -28,12 +30,11 @@ class HTTPHandler(MetricsHandler):
     """
 
     def __init__(self, *args, **kwargs):
-        self.docker_api = docker.APIClient()
+        self.docker_api: docker.APIClient = docker.APIClient()
         self.docker_client = docker.from_env()
         super().__init__(*args, **kwargs)
 
     # Override built-in method
-    # pylint: disable=invalid-name
     def do_GET(self):
         """
         Method to handle GET requests
@@ -48,7 +49,7 @@ class HTTPHandler(MetricsHandler):
 
             self._respond(200, 'OK')
 
-    def _respond(self, status, message):
+    def _respond(self, status: int, message: str):
         """
         Method to output a simple HTTP status and string to the client
         """
@@ -67,60 +68,110 @@ class HTTPHandler(MetricsHandler):
 
         registry = CollectorRegistry()
 
-        gauge = Gauge(
+        health_gauge = Gauge(
             'container_inspect_state_health_status',
             "Container's healthcheck value (binary)",
             labelnames=['id', 'name', 'value'],
-            registry=registry
+            registry=registry,
+        )
+        status_gauge = Gauge(
+            'container_inspect_state_running',
+            "Container's running state (binary)",
+            labelnames=['id', 'name'],
+            registry=registry,
+        )
+        started_at_gauge = Gauge(
+            'container_inspect_state_started_at',
+            "Container's start time (int)",
+            labelnames=['id', 'name'],
+            registry=registry,
+        )
+        exit_code_gauge = Gauge(
+            'container_inspect_state_exit_code',
+            "Container's exit code (int)",
+            labelnames=['id', 'name'],
+            registry=registry,
+        )
+        alert_threshold_gauge = Gauge(
+            'container_inspect_downtime_alert_threshold',
+            "Container's downtime alert threshold in seconds (int)",
+            labelnames=['id', 'name'],
+            registry=registry,
         )
 
-        for container in self.docker_client.containers.list():
+        for container in self.docker_client.containers.list(all=True):
             data = self.docker_api.inspect_container(container.id)
 
+            running: str = bool(data['State']['Running'])
+            started_at: int = data['State']['StartedAt']
+            exit_code: int = int(data['State']['ExitCode'])
+            alert_threshold = int(
+                data['Config']['Labels'].get('io.prometheus.alert.downtime', 3600)
+            )
+            starttime = numpy.datetime64(started_at)
+
+            status_gauge.labels(
+                container.id,
+                container.name,
+            ).set(int(running))
+            started_at_gauge.labels(container.id, container.name,).set(
+                int(int(starttime) / 1000000000)  # strip nanoseconds
+            )
+            exit_code_gauge.labels(
+                container.id,
+                container.name,
+            ).set(int(exit_code))
+            alert_threshold_gauge.labels(
+                container.id,
+                container.name,
+            ).set(alert_threshold)
+
             try:
-                health_str = data["State"]["Health"]["Status"]
-                label_values = [
+                health_str: str = data['State']['Health']['Status']
+                health_gauge.labels(
                     container.id,
                     container.name,
                     health_str,
-                ]
+                ).set(int(health_str == HEALTHY_STR))
             except KeyError:
                 pass
-            else:
-                gauge.labels(*label_values).set(int(health_str == HEALTHY_STR))
 
         self._respond(200, generate_latest(registry).decode())
 
 
-def healthy():
+def healthy() -> bool:
     """
     Simple funtion to return if all the requirements are met
     """
-    return all([
-        os.path.exists('/var/run/docker.sock'),
-    ])
+    return all(
+        [
+            os.path.exists('/var/run/docker.sock'),
+        ]
+    )
 
 
 if __name__ == '__main__':
-    def cli_parse():
+
+    def cli_parse() -> argparse.Namespace:
         """
         Function to parse the CLI
         """
         parser = argparse.ArgumentParser()
 
         parser.add_argument(
-            '-H', '--healthcheck',
+            '-H',
+            '--healthcheck',
             action='store_true',
             help='Simply exit with 0 for healthy or 1 when unhealthy',
         )
 
         return parser.parse_args()
 
-    def main():
+    def main() -> int:
         """
         main()
         """
-        args = cli_parse()
+        args: argparse.Namespace = cli_parse()
 
         if args.healthcheck:
             # Invert the sense of 'healthy' for Unix CLI usage

requirements.txt

@@ -1,3 +1,4 @@
 # To ensure app dependencies are ported from your virtual environment/host machine into your container, run 'pip freeze > requirements.txt' in the terminal to overwrite this file
 docker
+numpy
 prometheus_client