Feature Request: Access to {{ .Duration }} in alert message template

[thuck]

InfluxDB v1.1.1 (git: unknown unknown)
Kapacitor 1.3.2 (git: HEAD 2b6a049e5b9dc7d7e8d58a8bbde8d186dc8543e3)
Telegraf v1.3.1 (git: release-1.3 f93615672b02a41d9bc867bd92bf31c1d777989b)

Maybe I'm missing something, but If I try to use {{ .Duration }} as part of the alarm message I have an error:

ID: host_down
Error: alert5: template: message:1:95: executing "message" at <.Duration>: can't evaluate field Duration in type kapacitor.messageInfo
Template: 
Type: stream
Status: enabled
Executing: false
Created: 08 Oct 17 14:06 UTC
Modified: 08 Oct 17 19:02 UTC
LastEnabled: 08 Oct 17 19:02 UTC
Databases Retention Policies: ["telegraf"."autogen"]
TICKscript:
var data = stream
    |from()
        .database('telegraf')
        .retentionPolicy('autogen')
        .measurement('system')
        .groupBy('host')

data
    |deadman(1.0, 60s)
        .id('{{ index .Tags "host"}}/state')
        .message('{{ index .Tags "host" }} is {{ if eq .Level "OK" }} UP  {{ else }} DOWN  {{ end }} duration {{ .Duration }}')
        .details('')
        .stateChangesOnly(1m)
        .post('http://192.168.1.4:5000')
        .slack()
        .log('/tmp/cpu_alert_log.txt')

DOT:
digraph host_down {
stream0 -> from1;
from1 -> noop3;
stats2 -> derivative4;
derivative4 -> alert5;
}

Taking a look on the source code of the alert.go; I cannot find duration as part of the renderMessageAndDetails
Am I missing something or this is not implemented? Also if implemented how could I change this from nanoseconds to hours or minutes.

[desa]

@thuck What is Duration?

[thuck]

@desa for the AlertNode there is some predefined fields:

Each event that gets sent to a handler contains the following alert data:

    ID – the ID of the alert, user defined.
    Message – the alert message, user defined.
    Details – the alert details, user defined HTML content.
    Time – the time the alert occurred.
    Duration – the duration of the alert in nanoseconds.
    Level – one of OK, INFO, WARNING or CRITICAL.
    Data – influxql.Result containing the data that triggered the alert.

This is extract from here: https://docs.influxdata.com/kapacitor/v1.3//nodes/alert_node/

If you check the json generated by the alert there is the field for duration there:

{"id":"XXXX/state","message":"YYYYYY XXXX is  DOWN  ","details":"","time":"2017-10-10T21:28:00Z","duration":600000000000,"level":"CRITICAL","data":{"series":[{"name":"stats","tags":{"env":"YYYY","host":"XXXX","objectname":"System","project":"XXXX","role":"XXXXX"},"columns":["time","emitted"],"values":[["2017-10-10T21:28:00Z",0]]}]}}

So, as far as I can understand, the Duration here should be the duration that we have on the json, that's is related to the time that the alert is in that state; but if you try to use {{ .Duration }} as you would use {{ .Level }} it do not find it.

So this works:

.message('{{ .ID }} is {{ .Level }} value:{{ index .Fields "value" }}')

But this do not:

.message('{{ .ID }} is {{ .Level }} value:{{ index .Fields "value" }} time on state {{ .Duration }}')

[desa]

From the documentation it looks like the following are available in the message template

http://docs.influxdata.com/kapacitor/v1.3/nodes/alert_node/#message

Template for constructing a meaningful message for the alert. Available template data:
ID – The ID of the alert.
Name – Measurement name.
TaskName – The name of the task
Group – Concatenation of all group-by tags of the form [key=value,]+. If no groupBy is performed equal to literal 'nil'.
Tags – Map of tags. Use '{{ index .Tags "key" }}' to get a specific tag value.
Level – Alert Level, one of: INFO, WARNING, CRITICAL.
Fields – Map of fields. Use '{{ index .Fields "key" }}' to get a specific field value.
Time – The time of the point that triggered the event.

[desa]

Hmm weird, it should actually be available

https://github.com/influxdata/kapacitor/blob/master/alert/types.go#L100

// TemplateData is a structure containing all information available to use in templates for an Event.
type TemplateData struct {
	// The ID of the alert.
	ID string

	// The Message of the Alert
	Message string

	// Alert Level, one of: INFO, WARNING, CRITICAL.
	Level string

	// Time the event occurred.
	Time time.Time

	// Duration of the event
	Duration time.Duration

	// Measurement name
	Name string

	// Task name
	TaskName string

	// Concatenation of all group-by tags of the form [key=value,]+.
	// If not groupBy is performed equal to literal 'nil'
	Group string

	// Map of tags
	Tags map[string]string

	// Fields of alerting data point.
	Fields map[string]interface{}
}

[desa]

Wait, actually I'm wrong

https://github.com/influxdata/kapacitor/blob/master/alert.go#L1035-L1049

type messageInfo struct {
	idInfo

	// The ID of the alert.
	ID string

	// Fields of alerting data point.
	Fields map[string]interface{}

	// Alert Level, one of: INFO, WARNING, CRITICAL.
	Level string

	// Time
	Time time.Time
}

I'm going to change this to a Feature Request.

[thuck]

Great, thanks

[desa]

@thuck if you're interested, it'd be a pretty easy contribution. Essentially it'd be

Step 1

https://github.com/influxdata/kapacitor/blob/master/alert.go#L1035-L1049
Add Duration to the messageInfo struct

Step 2

https://github.com/influxdata/kapacitor/blob/master/alert.go#L698
Modify renderMessageAndDetails to take in a duration and set it on the messageInfo

Step 3

Write tests to verify the behavior.

[thuck]

@desa yep, I'll give a try on this, looks simple to implement