load.conf 2.3 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172
  1. # you can disable an alarm notification by setting the 'to' line to: silent
  2. # Calculate the base trigger point for the load average alarms.
  3. # This is the maximum number of CPU's in the system over the past 1
  4. # minute, with a special case for a single CPU of setting the trigger at 2.
  5. alarm: load_cpu_number
  6. on: system.load
  7. class: Utilization
  8. type: System
  9. component: Load
  10. os: linux
  11. hosts: *
  12. calc: ($active_processors == nan or $active_processors == 0) ? (nan) : ( ($active_processors < 2) ? ( 2 ) : ( $active_processors ) )
  13. units: cpus
  14. every: 1m
  15. info: Number of active CPU cores in the system
  16. # Send alarms if the load average is unusually high.
  17. # These intentionally _do not_ calculate the average over the sampled
  18. # time period because the values being checked already are averages.
  19. alarm: load_average_15
  20. on: system.load
  21. class: Utilization
  22. type: System
  23. component: Load
  24. os: linux
  25. hosts: *
  26. lookup: max -1m unaligned of load15
  27. calc: ($load_cpu_number == nan) ? (nan) : ($this)
  28. units: load
  29. every: 1m
  30. warn: ($this * 100 / $load_cpu_number) > (($status >= $WARNING) ? 175 : 200)
  31. delay: down 15m multiplier 1.5 max 1h
  32. summary: Host load average (15 minutes)
  33. info: System load average for the past 15 minutes
  34. to: silent
  35. alarm: load_average_5
  36. on: system.load
  37. class: Utilization
  38. type: System
  39. component: Load
  40. os: linux
  41. hosts: *
  42. lookup: max -1m unaligned of load5
  43. calc: ($load_cpu_number == nan) ? (nan) : ($this)
  44. units: load
  45. every: 1m
  46. warn: ($this * 100 / $load_cpu_number) > (($status >= $WARNING) ? 350 : 400)
  47. delay: down 15m multiplier 1.5 max 1h
  48. summary: System load average (5 minutes)
  49. info: System load average for the past 5 minutes
  50. to: silent
  51. alarm: load_average_1
  52. on: system.load
  53. class: Utilization
  54. type: System
  55. component: Load
  56. os: linux
  57. hosts: *
  58. lookup: max -1m unaligned of load1
  59. calc: ($load_cpu_number == nan) ? (nan) : ($this)
  60. units: load
  61. every: 1m
  62. warn: ($this * 100 / $load_cpu_number) > (($status >= $WARNING) ? 700 : 800)
  63. delay: down 15m multiplier 1.5 max 1h
  64. summary: System load average (1 minute)
  65. info: System load average for the past 1 minute
  66. to: silent