more elastic alarm thresholds for cpu, ram and swap

diff --git a/conf.d/health.d/cpu.conf b/conf.d/health.d/cpu.conf
index 4d79fc799a0fcd2a56df102f6e19ed66145dd602..2b04c794d0919a203fe5ce6af204d785e4b256c2 100644 (file)
--- a/conf.d/health.d/cpu.conf
+++ b/conf.d/health.d/cpu.conf
@@ -4,8 +4,8 @@ template: 10min_cpu_usage
   lookup: average -10m unaligned of user,system,nice,softirq,irq,guest,guest_nice
    units: %
    every: 1m
-    warn: $this > (($status >= $WARNING)  ? (70) : (80))
-    crit: $this > (($status == $CRITICAL) ? (80) : (90))
+    warn: $this > (($status >= $WARNING)  ? (75) : (85))
+    crit: $this > (($status == $CRITICAL) ? (85) : (95))
    delay: down 15m multiplier 1.5 max 1h
     info: average cpu utilization for the last 10 minutes
       to: sysadmin
@@ -15,8 +15,8 @@ template: 10min_cpu_iowait
   lookup: average -10m unaligned of iowait
    units: %
    every: 1m
-    warn: $this > (($status >= $WARNING)  ? (5)  : (10))
-    crit: $this > (($status == $CRITICAL) ? (20) : (30))
+    warn: $this > (($status >= $WARNING)  ? (20) : (40))
+    crit: $this > (($status == $CRITICAL) ? (40) : (50))
    delay: down 15m multiplier 1.5 max 1h
     info: average CPU wait I/O for the last 10 minutes
       to: sysadmin

diff --git a/conf.d/health.d/ram.conf b/conf.d/health.d/ram.conf
index 216b82fed853517d9d1b97c4d820f16e1e80f4e1..d60df75b2040e8f6e4054d540b6be291095cd167 100644 (file)
--- a/conf.d/health.d/ram.conf
+++ b/conf.d/health.d/ram.conf
@@ -4,8 +4,8 @@
     calc: $used * 100 / ($used + $cached + $free)
    units: %
    every: 10s
-    warn: $this > (($status >= $WARNING)  ? (70) : (80))
-    crit: $this > (($status == $CRITICAL) ? (80) : (90))
+    warn: $this > (($status >= $WARNING)  ? (80) : (90))
+    crit: $this > (($status == $CRITICAL) ? (90) : (98))
    delay: down 15m multiplier 1.5 max 1h
     info: system RAM usage
       to: sysadmin

diff --git a/conf.d/health.d/swap.conf b/conf.d/health.d/swap.conf
index 98ba9b52f2a077dd5efb7379987137859b210ceb..7f57560e245f6d75bc2d57cfb663ecd1d35c5e0e 100644 (file)
--- a/conf.d/health.d/swap.conf
+++ b/conf.d/health.d/swap.conf
@@ -6,8 +6,8 @@
     calc: $this / 1024 * 100 / ( $system.ram.used + $system.ram.cached + $system.ram.free )
    units: % of RAM
    every: 1m
-    warn: $this > (($status >= $WARNING)  ? (5)  : (10))
-    crit: $this > (($status == $CRITICAL) ? (15) : (20))
+    warn: $this > (($status >= $WARNING)  ? (10) : (20))
+    crit: $this > (($status == $CRITICAL) ? (20) : (30))
    delay: up 0 down 15m multiplier 1.5 max 1h
     info: the amount of memory swapped in the last 30 minutes, as a percentage of the system RAM
       to: sysadmin
@@ -28,8 +28,8 @@
     calc: $used * 100 / ( $used + $free )
    units: %
    every: 10s
-    warn: $this > (($status >= $WARNING)  ? (60) : (80))
-    crit: $this > (($status == $CRITICAL) ? (80) : (90))
+    warn: $this > (($status >= $WARNING)  ? (80) : (90))
+    crit: $this > (($status == $CRITICAL) ? (90) : (98))
    delay: up 0 down 15m multiplier 1.5 max 1h
     info: the percentage of swap memory used
       to: sysadmin