From c47a05aa396381ec46340cfeb8360e9bd3b17354 Mon Sep 17 00:00:00 2001
From: Michael Osthege <michael.osthege@outlook.com>
Date: Sun, 12 May 2024 12:49:32 +0200
Subject: [PATCH 1/3] Add `refresh_age` metric

---
 exporter.py | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/exporter.py b/exporter.py
index 51ddd43..fe829f1 100755
--- a/exporter.py
+++ b/exporter.py
@@ -152,6 +152,7 @@ def run(self):
             try:
                 with generate_metrics_summary.labels(self.ccu_host).time():
                     self.generate_metrics()
+                    self.refresh_time = time.time()
             except OSError as os_error:
                 logging.info("Failed to generate metrics: {0}".format(os_error))
                 error_counter.labels(self.ccu_host).inc()
@@ -183,6 +184,11 @@ def __init__(self, ccu_host, ccu_port, auth, gathering_interval, reload_names_in
         self.gathering_interval = int(gathering_interval)
         self.reload_names_interval = int(reload_names_interval)
         self.devicecount = Gauge('devicecount', 'Number of processed/supported devices', labelnames=['ccu'], namespace=self.METRICS_NAMESPACE)
+        # Upon request export the seconds since the last successful update.
+        # This is robust against internal crashes and can be used by the healthcheck.
+        self.refresh_time = time.time()
+        self.refresh_age = Gauge("refresh_age_seconds", "Seconds since the last successful refresh.", labelnames=["ccu"], namespace=self.METRICS_NAMESPACE)
+        self.refresh_age.labels(self.ccu_host).set_function(lambda: time.time() - self.refresh_time)
 
     def generate_metrics(self):
         logging.info("Gathering metrics")

From 463a2e8f16e1e31028c7e5058bb5ee00a880a577 Mon Sep 17 00:00:00 2001
From: Michael Osthege <michael.osthege@outlook.com>
Date: Sun, 12 May 2024 12:53:51 +0200
Subject: [PATCH 2/3] Add a healthcheck

---
 Dockerfile     |  3 +++
 healthcheck.sh | 14 ++++++++++++++
 2 files changed, 17 insertions(+)
 create mode 100644 healthcheck.sh

diff --git a/Dockerfile b/Dockerfile
index 83039dc..63441cc 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -3,7 +3,10 @@ COPY requirements.txt /tmp
 RUN pip3 install --no-cache-dir -r /tmp/requirements.txt
 
 COPY exporter.py /usr/local/bin/homematic_exporter
+COPY healthcheck.sh /usr/local/bin/healthcheck.sh
 
 ENTRYPOINT [ "/usr/local/bin/homematic_exporter" ]
 
 EXPOSE 8010
+HEALTHCHECK --interval=20s --timeout=3s \
+    CMD bash /usr/local/bin/healthcheck.sh
diff --git a/healthcheck.sh b/healthcheck.sh
new file mode 100644
index 0000000..4d4c84e
--- /dev/null
+++ b/healthcheck.sh
@@ -0,0 +1,14 @@
+#!/bin/bash
+
+# First argument is the maximum age in seconds (default: 60)
+maxage=${1:-60}
+
+# Get the age of the last successful metrics refresh in seconds
+age=$(curl -s http://localhost:9040/metrics | grep 'homematic_refresh_age{' | cut -d ' ' -f2 | cut -d '.' -f1)
+
+if [[ $age -lt $maxage ]]; then
+    exit 0
+else
+    # Maximum age exceeded → unhealthy
+    exit 1
+fi

From 36c4135f1508a9419b52091ebb8625ac85714503 Mon Sep 17 00:00:00 2001
From: Michael Osthege <michael.osthege@outlook.com>
Date: Wed, 22 May 2024 01:08:01 +0200
Subject: [PATCH 3/3] Include curl to fix healthcheck

---
 Dockerfile     | 1 +
 healthcheck.sh | 2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/Dockerfile b/Dockerfile
index 63441cc..7d10f3d 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -1,5 +1,6 @@
 FROM python:3.12-slim-bookworm
 COPY requirements.txt /tmp
+RUN apt-get update && apt-get install curl -y
 RUN pip3 install --no-cache-dir -r /tmp/requirements.txt
 
 COPY exporter.py /usr/local/bin/homematic_exporter
diff --git a/healthcheck.sh b/healthcheck.sh
index 4d4c84e..4834867 100644
--- a/healthcheck.sh
+++ b/healthcheck.sh
@@ -4,7 +4,7 @@
 maxage=${1:-60}
 
 # Get the age of the last successful metrics refresh in seconds
-age=$(curl -s http://localhost:9040/metrics | grep 'homematic_refresh_age{' | cut -d ' ' -f2 | cut -d '.' -f1)
+age=$(curl -s http://localhost:9040/metrics | grep 'homematic_refresh_age_seconds{' | cut -d ' ' -f2 | cut -d '.' -f1)
 
 if [[ $age -lt $maxage ]]; then
     exit 0