Merge branch 'alloy' into smartctl

.env.sample

@@ -30,6 +30,9 @@ LIVE_DEBUGGING=false
 # server is remote
 # PROMETHEUS_REMOTE_WRITE_URL=https://prometheus.$DOMAIN/api/v1/write
 
+# Monitor physical disks health
+# COMPOSE_FILE="$COMPOSE_FILE:compose.smartctl.yml"
+
 # Monitoring Server
 #
 ## Prometheus
@@ -99,9 +102,3 @@ LIVE_DEBUGGING=false
 
 # Node memory usage alert will trigger when memory usage is above the given number in percent
 #ALERT_NODE_MEMORY_USAGE=85
-
-# Tell Traefik to keep access logs (could be very verbose)
-ALLOY_ACCESS_LOGS=false
-PROMETHEUS_ACCESS_LOGS=false
-LOKI_ACCESS_LOGS=false
-GRAFANA_ACCESS_LOGS=false

README.md

@@ -138,3 +138,11 @@ It is possible to enable the following alerts, by uncommenting the corresponding
 
 - node disk space: `ALERT_NODE_DISK_SPACE_LEFT`
 - node memory usage: `ALERT_NODE_MEMORY_USAGE`
+
+## smart monitoring
+
+To be able monitor hard drive health data, you need to configure
+`smartd` to run on the host system, and also the
+`collect-smartctl-json.sh` script provided here (via cronjob or as
+a `smartd` hook).  This is a limitation on Docker Swarm, which prevents
+the `smartctl_exporter` from running on privileged mode.

collect-smartctl-json.service

@@ -0,0 +1,6 @@
+[Unit]
+Description=Collect SMART data
+
+[Service]
+Type=oneshot
+ExecStart=/usr/local/bin/collect-smartctl-json.sh

collect-smartctl-json.sh

@@ -0,0 +1,69 @@
+#! /bin/bash
+# Adapted from https://github.com/prometheus-community/smartctl_exporter/blob/master/collect-smartctl-json.sh
+
+script_dir=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )
+
+# Data directory to dump smartctl output
+# This directory will be created if it doesn't exist
+data_dir="/var/lib/smartmontools/json"
+
+# The original script used --xall but that doesn't work
+# This matches the command in readSMARTctl()
+smartctl_args="--json --info --health --attributes --tolerance=verypermissive \
+--nocheck=standby --format=brief --log=error"
+
+# Ignore this devices
+smartctl_ignore_dev_regex="^(/dev/bus)"
+
+# Determine the json query tool to use
+if command -v jq >/dev/null; then
+	json_tool="jq"
+	json_args="--raw-output"
+elif command -v yq >/dev/null; then
+	json_tool="yq"
+	json_args="--unwrapScalar"
+else
+	echo -e "One of 'yq' or 'jq' is required. Please try again after \
+installing one of them"
+	exit 1
+fi
+
+if [[ ! "${UID}" -eq 0 ]] && ! command -v sudo >/dev/null; then
+	# Not root and sudo doesn't exist
+	echo "sudo does not exist. Please run this as root"
+	exit 1
+fi
+
+SUDO="sudo"
+if [[ "${UID}" -eq 0 ]]; then
+	# Don't use sudo if root
+	SUDO=""
+fi
+
+[[ ! -d "${data_dir}" ]] && mkdir --parents "${data_dir}"
+
+if [[ $# -ne 0 ]]; then
+	devices="${1}"
+else
+	devices="$(smartctl --scan --json | "${json_tool}" "${json_args}" \
+".devices[].name | select(test(\"${smartctl_ignore_dev_regex}\") | not)")"
+  mapfile -t devices <<< "${devices[@]}"
+fi
+
+for device in "${devices[@]}"
+  do
+	echo -n "Collecting data for '${device}'..."
+	# shellcheck disable=SC2086
+	data="$($SUDO smartctl ${smartctl_args} ${device})"
+	# Accommodate a smartmontools pre-7.3 bug
+	data=${data#"  Pending defect count:"}
+	type="$(echo "${data}" | "${json_tool}" "${json_args}" '.device.type')"
+	family="$(echo "${data}" | "${json_tool}" "${json_args}" \
+'select(.model_family != null) | .model_family | sub(" |/" ; "_" ; "g")
+ | sub("\"|\\(|\\)" ; "" ; "g")')"
+	model="$(echo "${data}" | "${json_tool}" "${json_args}" \
+'.model_name | sub(" |/" ; "_" ; "g") | sub("\"|\\(|\\)" ; "" ; "g")')"
+	device_name="$(basename "${device}")"
+	echo -e "\tSaving to ${device_name}.json"
+	echo "${data}" > "${data_dir}/${device_name}.json"
+done

collect-smartctl-json.timer

@@ -0,0 +1,9 @@
+[Unit]
+Description=Collect SMART data
+
+[Timer]
+OnCalendar=hourly
+Persistent=true
+
+[Install]
+WantedBy=timers.target

compose.grafana.yml

@@ -43,7 +43,6 @@ services:
         - "traefik.http.routers.${STACK_NAME}-grafana.entrypoints=web-secure"
         - "traefik.http.routers.${STACK_NAME}-grafana.tls=true"
         - "traefik.http.routers.${STACK_NAME}-grafana.tls.certresolver=${LETS_ENCRYPT_ENV}"
-        - "traefik.http.routers.${STACK_NAME}-grafana.observability.accesslogs=${GRAFANA_ACCESS_LOGS:-false}"
     healthcheck:
       test: "wget -q http://localhost:3000/healthz -O/dev/null"
       interval: 5s

compose.loki.yml

@@ -34,7 +34,6 @@ services:
         - "traefik.http.routers.${STACK_NAME}-loki.tls=true"
         - "traefik.http.routers.${STACK_NAME}-loki.tls.certresolver=${LETS_ENCRYPT_ENV}"
         - "traefik.http.routers.${STACK_NAME}-loki.middlewares=basicauth@file"
-        - "traefik.http.routers.${STACK_NAME}-loki.observability.accesslogs=${LOKI_ACCESS_LOGS:-false}"
 
 
 configs:

compose.prometheus.yml

@@ -32,7 +32,6 @@ services:
         - "traefik.http.routers.${STACK_NAME}-prometheus.tls=true"
         - "traefik.http.routers.${STACK_NAME}-prometheus.tls.certresolver=${LETS_ENCRYPT_ENV}"
         - "traefik.http.routers.${STACK_NAME}-prometheus.middlewares=basicauth@file"
-        - "traefik.http.routers.${STACK_NAME}-prometheus.observability.accesslogs=${PROMETHEUS_ACCESS_LOGS:-false}"
 
 configs:
   prometheus_yml:

compose.smartctl.yml

@@ -0,0 +1,18 @@
+---
+version: "3.8"
+services:
+  smartctl:
+    image: "prometheuscommunity/smartctl-exporter:v0.14.0"
+    volumes:
+    - "/dev:/dev"
+    - "/var/lib/smartmontools/json:/debug"
+    command:
+    - "--smartctl.fake-data"
+    - "--smartctl.interval=1h"
+    networks:
+    - "proxy"
+    deploy:
+      labels:
+        - "prometheus.io/scrape=true"
+        - "prometheus.io/port=9633"
+        - "prometheus.io/path=/metrics"

compose.yml

@@ -38,7 +38,6 @@ services:
         - "traefik.http.routers.${STACK_NAME}-alloy.tls=true"
         - "traefik.http.routers.${STACK_NAME}-alloy.tls.certresolver=${LETS_ENCRYPT_ENV}"
         - "traefik.http.routers.${STACK_NAME}-alloy.middlewares=basicauth@file"
-        - "traefik.http.routers.${STACK_NAME}-alloy.observability.accesslogs=${ALLOY_ACCESS_LOGS:-false}"
 configs:
   config_alloy:
     template_driver: golang