From 9f2025235d888eb4a7822024a5fad2e288388814 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Isma=C3=ABl=20Bouya?= <ismael.bouya@normalesup.org>
Date: Sun, 1 Dec 2019 18:25:16 +0100
Subject: Add monitoring for backup-2

---
 .../monitoring/plugins/check_last_file_date        | 26 ++++++++++++++++
 .../monitoring/plugins/check_postgres_replication  | 35 ++++++++++++++++++++++
 2 files changed, 61 insertions(+)
 create mode 100755 modules/private/monitoring/plugins/check_last_file_date
 create mode 100755 modules/private/monitoring/plugins/check_postgres_replication

(limited to 'modules/private/monitoring/plugins')

diff --git a/modules/private/monitoring/plugins/check_last_file_date b/modules/private/monitoring/plugins/check_last_file_date
new file mode 100755
index 0000000..df45bbc
--- /dev/null
+++ b/modules/private/monitoring/plugins/check_last_file_date
@@ -0,0 +1,26 @@
+#!/bin/bash
+ 
+STATE_OK=0
+STATE_WARNING=1
+STATE_CRITICAL=2
+STATE_UNKNOWN=3
+  
+base_path=$1
+hours=$2
+
+last_date=$(find $base_path -mindepth 1 -maxdepth 1 -printf "%T@\n" 2>/dev/null | sort | tail -n 1)
+
+if [ -z "$last_date" ]; then
+  echo "UNKNOWN: Could not read folder"
+  exit $STATE_UNKNOWN
+else
+  LC_ALL=C last_date=$(printf "%.*f" 0 $last_date)
+  min_date=$(date -d "$hours hours ago" "+%s")
+  if [ "$min_date" -lt "$last_date" ]; then
+    echo "OK: Last file $(date -d @$last_date)"
+    exit $STATE_OK
+  else
+    echo "CRITICAL: Last file $(date -d @$last_date)"
+    exit $STATE_CRITICAL
+  fi
+fi
diff --git a/modules/private/monitoring/plugins/check_postgres_replication b/modules/private/monitoring/plugins/check_postgres_replication
new file mode 100755
index 0000000..009b4d5
--- /dev/null
+++ b/modules/private/monitoring/plugins/check_postgres_replication
@@ -0,0 +1,35 @@
+#!/bin/bash
+
+STATE_OK=0
+STATE_WARNING=1
+STATE_CRITICAL=2
+STATE_UNKNOWN=3
+
+user=$1
+host=$2
+port=$3
+
+lag=$(psql -h $host -p $port -A -t -c "SELECT COALESCE(EXTRACT(EPOCH FROM replay_lag),0) FROM pg_stat_replication WHERE usename='$user'" 2>/dev/null)
+exit_code=$?
+
+if [[ $exit_code -ne 0 ]]; then
+  echo "UNKNOWN - Impossible to run psql command"
+  exit $STATE_UNKNOWN
+elif [[ -z "$lag" ]]; then
+  echo "UNKNOWN - No replication found for $user"
+  exit $STATE_UNKNOWN
+else
+  output="Replication lag for $user is ${lag}s"
+  LC_ALL=C lag=$(printf "%.*f" 0 $lag)
+
+  if [[ $lag -lt 5 ]]; then
+    echo "OK - $output"
+    exit $STATE_OK
+  elif [[ $lag -lt 10 ]]; then
+    echo "WARNING - $output"
+    exit $STATE_WARNING
+  else
+    echo "CRITICAL - $output"
+    exit $STATE_CRITICAL
+  fi
+fi
-- 
cgit v1.2.3