regel

data_dir = "/var/lib/kapacitor"

[[slack]]
  enabled = true
  default = true
  workspace = "your_workspace"
  url = "your_url"
  channel = "your_channel"
  global = false
  state-changes-only = false

regel

var model

var from_measurement

var out_db

var retention_policy

var out_measurement

regel

var model = 'telegraf_metrics_count_value_10s'

var from_measurement = 'loudml'

var out_db = 'telegraf'

var retention_policy = 'autogen'

var out_measurement = 'predicted'

regel

git clone https://github.com/regel/loudml
cd loudml/docker/compose
vim etc/kapacitor.conf # notifications settings
vim etc/loudml.yml # optional: database settings

regel

// the machine learning model name, as defined in settings
var model = 'telegraf_metrics_count_value_10s'
// the measurement name, used by Loud ML to output predictions; default = loudml
var from_measurement = 'loudml'
// the output database name for output data written by this pipeline
var out_db = 'telegraf'
// the retention policy; let's use the default
var retention_policy = 'autogen'
// the output measurement name for output data written by this pipeline
var out_measurement = 'predicted'

regel

var data = stream
    // this from node selects all input data received in the given measurement
    |from()
        .measurement(from_measurement)
        .where(lambda: "model" == model)
    // this eval node remembers the model name
    |eval(lambda: model)
        .as('model')
        .keep()
    // the next five lines save the data to the target database into a new measurement

regel

var pos = data
    // the next node counts abnormal data points
    |stateCount(lambda: "is_anomaly" == TRUE)
    |alert()
    // Warn after 1 point
        .warn(lambda: "state_count" >= 1)
    // Critical after 5 points and abnormal higher than 90 / 100
        .crit(lambda: "state_count" >= 5 AND "score" > 90.0)
        .message('{{ .Time }}: Hey, unexpected situation detected by model={{ index .Fields "model" }} state=ongoing score={{ index .Fields "score" | printf "%0.3f" }}.')
        .slack()

regel

var neg = data
    |stateCount(lambda: "is_anomaly" == FALSE)
    |alert()
        .info(lambda: "state_count" == 1)
        .message('{{ .Time }}: situation back to normal model={{ index .Fields "model" }} state=finished score={{ index .Fields "score" | printf "%0.3f" }}.')
        .slack()

regel

data_dir = "/var/lib/kapacitor"

[pagerduty2]
  enabled = true
  routing-key = ""
  url = "https://events.pagerduty.com/v2/enqueue"
  global = false
  
[logging]
    # Destination for logs

regel

var pos = data
    |changeDetect('is_anomaly')
    |alert()
        .warn(lambda: "is_anomaly" == TRUE)
        .crit(lambda: "is_anomaly" == TRUE AND "score" > 90.0)
        .message('{{if ne .Level "OK" }} Hey, unexpected situation detected score={{ index .Fields "score" | printf "%0.3f" }} {{else}} Back to normal {{end}}')
        .pagerDuty2()