| 1 | |
| 2 | # you can disable an alarm notification by setting the 'to' line to: silent |
| 3 | |
| 4 | template: 10min_cpu_usage |
| 5 | on: system.cpu |
| 6 | class: Utilization |
| 7 | type: System |
| 8 | component: CPU |
| 9 | host labels: _os=linux |
| 10 | lookup: average -10m unaligned of user,system,softirq,irq,guest |
| 11 | units: % |
| 12 | every: 1m |
| 13 | warn: $this > (($status >= $WARNING) ? (75) : (85)) |
| 14 | crit: $this > (($status == $CRITICAL) ? (85) : (95)) |
| 15 | delay: down 15m multiplier 1.5 max 1h |
| 16 | summary: System CPU utilization |
| 17 | info: Average CPU utilization over the last 10 minutes (excluding iowait, nice and steal) |
| 18 | to: sysadmin |
| 19 | |
| 20 | template: 10min_cpu_iowait |
| 21 | on: system.cpu |
| 22 | class: Utilization |
| 23 | type: System |
| 24 | component: CPU |
| 25 | host labels: _os=linux |
| 26 | lookup: average -10m unaligned of iowait |
| 27 | units: % |
| 28 | every: 1m |
| 29 | warn: $this > (($status >= $WARNING) ? (20) : (40)) |
| 30 | delay: up 30m down 30m multiplier 1.5 max 2h |
| 31 | summary: System CPU iowait time |
| 32 | info: Average CPU iowait time over the last 10 minutes |
| 33 | to: silent |
| 34 | |
| 35 | template: 20min_steal_cpu |
| 36 | on: system.cpu |
| 37 | class: Latency |
| 38 | type: System |
| 39 | component: CPU |
| 40 | host labels: _os=linux |
| 41 | lookup: average -20m unaligned of steal |
| 42 | units: % |
| 43 | every: 5m |
| 44 | warn: $this > (($status >= $WARNING) ? (5) : (10)) |
| 45 | delay: down 1h multiplier 1.5 max 2h |
| 46 | summary: System CPU steal time |
| 47 | info: Average CPU steal time over the last 20 minutes |
| 48 | to: silent |
| 49 | |
| 50 | ## Windows |
| 51 | |
| 52 | template: 10min_cpu_usage |
| 53 | on: system.cpu |
| 54 | class: Utilization |
| 55 | type: System |
| 56 | component: CPU |
| 57 | host labels: _os=windows |
| 58 | lookup: average -10m unaligned of irq,user,system,dpc |
| 59 | units: % |
| 60 | every: 1m |
| 61 | warn: $this > (($status >= $WARNING) ? (75) : (85)) |
| 62 | crit: $this > (($status == $CRITICAL) ? (85) : (95)) |
| 63 | delay: up 1m down 15m multiplier 1.5 max 1h |
| 64 | summary: System CPU utilization |
| 65 | info: Average CPU utilization over the last 10 minutes |
| 66 | to: sysadmin |
| 67 | |
| 68 | ## FreeBSD |
| 69 | template: 10min_cpu_usage |
| 70 | on: system.cpu |
| 71 | class: Utilization |
| 72 | type: System |
| 73 | component: CPU |
| 74 | host labels: _os=freebsd |
| 75 | lookup: average -10m unaligned of user,system,interrupt |
| 76 | units: % |
| 77 | every: 1m |
| 78 | warn: $this > (($status >= $WARNING) ? (75) : (85)) |
| 79 | crit: $this > (($status == $CRITICAL) ? (85) : (95)) |
| 80 | delay: down 15m multiplier 1.5 max 1h |
| 81 | summary: System CPU utilization |
| 82 | info: Average CPU utilization over the last 10 minutes (excluding nice) |
| 83 | to: sysadmin |