137 lines
3.9 KiB
HCL
137 lines
3.9 KiB
HCL
# Monitor CPU Utilisation
|
|
resource "datadog_monitor" "cpumonitor" {
|
|
name = "cpu monitor"
|
|
type = "metric alert"
|
|
message = "CPU usage alert"
|
|
query = "avg(last_1m):avg:system.cpu.system{*} by {host} > 60"
|
|
monitor_thresholds {
|
|
ok = 20
|
|
warning = 50
|
|
critical = 60
|
|
}
|
|
}
|
|
|
|
# Monitor Memory Utilisation
|
|
resource "datadog_monitor" "memorymonitor" {
|
|
name = "Usable Memory"
|
|
type = "query alert"
|
|
evaluation_delay = "15"
|
|
include_tags = true
|
|
locked = false
|
|
message = "Plaaper de plaaper de plaap"
|
|
new_host_delay = 300
|
|
no_data_timeframe = 0
|
|
#notify_audit = 0
|
|
#notify_no_data = 0
|
|
priority = 0
|
|
renotify_interval = 0
|
|
require_full_window = true
|
|
#restricted_roles = []
|
|
timeout_h = 0
|
|
query = "max(last_5m):avg:system.mem.usable{*} by {host} / avg:system.mem.total{*} by {host} * 100 < 5"
|
|
|
|
monitor_thresholds {
|
|
critical = 5
|
|
warning = 10
|
|
}
|
|
}
|
|
|
|
# Monitor System Load
|
|
resource "datadog_monitor" "systemload" {
|
|
name = "System Load"
|
|
type = "query alert"
|
|
evaluation_delay = "15"
|
|
include_tags = true
|
|
locked = false
|
|
new_host_delay = 300
|
|
no_data_timeframe = 0
|
|
#notify_audit = false
|
|
priority = 0
|
|
renotify_interval = 0
|
|
require_full_window = true
|
|
#restricted_roles = []
|
|
tags = []
|
|
timeout_h = 0
|
|
message = "System Load exceeding set Thresholg, please investigate"
|
|
query = "min(last_30m):( avg:system.load.norm.5{*} by {host} ) > 2.5"
|
|
|
|
monitor_thresholds {
|
|
critical = "2.5"
|
|
warning = "2"
|
|
}
|
|
|
|
}
|
|
# Monitor Disk Utilisation
|
|
resource "datadog_monitor" "disk_usage" {
|
|
name = "Disk Space"
|
|
type = "query alert"
|
|
evaluation_delay = 15
|
|
include_tags = true
|
|
locked = false
|
|
new_host_delay = 300
|
|
no_data_timeframe = 0
|
|
#notify_audit = 0
|
|
priority = 0
|
|
renotify_interval = 0
|
|
require_full_window = true
|
|
#restricted_roles = []
|
|
tags = []
|
|
timeout_h = 0
|
|
message = "Disk Usage is exceeding set threshold, please investigate"
|
|
query = "max(last_5m):avg:system.disk.in_use{*} by {host,device} * 100 > 90"
|
|
|
|
monitor_thresholds {
|
|
critical = "90"
|
|
warning = "80"
|
|
}
|
|
}
|
|
|
|
# Monitor Disk Inode Usage
|
|
resource "datadog_monitor" "disk_inodes" {
|
|
name = "Disk Inodes Usage"
|
|
type = "query alert"
|
|
evaluation_delay = 15
|
|
include_tags = true
|
|
locked = false
|
|
new_host_delay = 300
|
|
no_data_timeframe = 0
|
|
#notify_audit = 0
|
|
priority = 0
|
|
renotify_interval = 0
|
|
require_full_window = true
|
|
#restricted_roles = []
|
|
tags = []
|
|
timeout_h = 0
|
|
message = "Disk Inode Usage is exceeding set threshold, please investigate"
|
|
query = "min(last_5m):avg:system.fs.inodes.in_use{*} by {host,device} * 100 > 95"
|
|
|
|
monitor_thresholds {
|
|
critical = "95"
|
|
warning = "90"
|
|
}
|
|
}
|
|
|
|
# Monitor Disk Space Forecast
|
|
resource "datadog_monitor" "disk_forecast" {
|
|
name = "Disk Usage Forecast"
|
|
type = "query alert"
|
|
evaluation_delay = 15
|
|
include_tags = true
|
|
locked = false
|
|
new_host_delay = 300
|
|
no_data_timeframe = 0
|
|
#notify_audit = 0
|
|
priority = 0
|
|
renotify_interval = 0
|
|
require_full_window = true
|
|
#restricted_roles = []
|
|
tags = []
|
|
timeout_h = 0
|
|
message = "Disk Usage is exceeding set threshold, please investigate"
|
|
query = "max(next_1w):forecast(avg:system.disk.in_use{*} by {host,device} * 100, 'linear', 1, interval='60m', history='1w', model='default') >= 80"
|
|
|
|
monitor_thresholds {
|
|
critical = "80"
|
|
warning = "72"
|
|
}
|
|
} |