Custom metrics, traces and logs (BPM)

You can add custom metrics, traces and logs in the workflow DSL at each task level. Once you add these BPM (Business performance metrics) then these will be available out of the box along with APM (Application performance metrics).

DSL spec for custom metrics

You can specify custom metrics in a task by follwing the below DSL spec. Check Prometheis metrics types to find the types of available metrics.

metrics:
- name: metric_name
  type: counter|gauge|histogram|summary
  labels: 
    label1: val1
    label2: val2
          
  # following functions depending on the metric type and all of them could be scripts, can use inputs/outputs
  inc: 10
  dec: 10
  set: 100
  observe: 2000
  timer: true|false(boolean) starts at the beginning of workflow/task and ends at the end of workflow/task

metric type	function
counter	inc
gauge	inc, dec, set
histogram	observe, timer
summary	observe

Refer prom-client to know more about the functions to be used for each metrics type.

Example

In the following example, we are using two custom metrics:
1. httpbin_calls_total: counter type metric, counter is incremented by 1.
2. httpbin_calls_duration: histogram type metric, timer is set to true to record duration.

summary: Call an API and transform the 
tasks:
    - id: httpbin_step1 # the response of this will be accessible within the parent step key, under the step1 sub key
      name: http bin step
      description: Hit http bin with some dummy data. It will send back same as response
      fn: datasource.api.post./anything       
      metrics:
        - name: httpbin_calls_total
          help: 'httpbin_calls_total counter of httpbin requests labeled with: method, status_code'
          type: counter
          labels:
            method: httpbin
            status_code: <% outputs.httpbin_step1.code %>               
          inc: 1
        - name: httpbin_calls_duration
          help: 'httpbin_calls_duration duration histogram of httpbin responses labeled with: method, status_code'
          type: histogram
          labels:
            method: httpbin
            status_code: <% outputs.httpbin_step1.code %>               
          timer: true          
      args:
        params: <% inputs.query %>
        data: <% inputs.body %>

DSL spec for custom trace

You can add custom trace in the workflow tasks. Then these custom traces will be traced along with the default traces. Check OTel traces to understand more about trace, span, attributes, etc.,

trace:
  name: span_name
  attributes:
    attribute1: value1
    attribute2: value2

Example

In the following example, we are creating a new span named httpbin_trace with span attributes request and param. This span gets created when the task starts and ended when the task completes its execution.

summary: Call an API and transform the 
tasks:
  - id: httpbin_step1 # the response of this will be accessible within the parent step key, under the step1 sub key
    name: http bin step
    description: Hit http bin with some dummy data. It will send back same as response
    fn: datasource.api.post./anything
    trace:
      name: httpbin_trace
      attributes:
          request: <%inputs.body%>
          param: <%inputs.query%>
    args:
      params: <% inputs.query %>
      data: <% inputs.body %>

DSL spec for custom logs

You can add custom log before and after a task in a workflow.

logs:
  before:
    level: fatal|error|warn|info|debug|trace # refer pino for levels
    message: 'Sample log before'
    params: 
      param1: val1
      param2: val2
    attributes:
      request:
        query: <%inputs.query%>
  after:
    level: info
    message: 'Sample log after'
    params:
    attributes: 

Please refer to OTEL Logging Data model for understanding of fields dumped in the logs.

message and params are part of Body field and attributes are part of Attributes field in the log.

Example

In the following example, we are adding two additional logs before and after the task execution.

summary: Call an API and transform the 
tasks:
    - id: httpbin_step1 # the response of this will be accessible within the parent step key, under the step1 sub key
      name: http bin step
      description: Hit http bin with some dummy data. It will send back same as response
      fn: datasource.api.post./anything
      logs:
        before:
          level: error
          message: 'Hello'
          params: 
            - key1: v1
              key2: v2
            - v1
          attributes: 
            request: <%inputs.query%>
        after:
          level: error
          message: 'World'
          params: 
            key1: v1
            key2: v2
          attributes: 
            customer_name: <% outputs.httpbin_step1.data.json.customer_name %> 
      args:
        params: <% inputs.query %>
        data: <% inputs.body %>

Sample Logs

{"Body":"Hello [{\"key1\":\"v1\",\"key2\":\"v2\"},\"v1\"]","Timestamp":"2024-04-10T09:40:45.122Z000000","SeverityNumber":9,"SeverityText":"INFO","TraceId":"afde0bf5bb3533d932c1c04c30d91172","SpanId":"ad477b2cf81ca711","TraceFlags":"01","Resource":{"service.name":"unknown_service:node","host.hostname":"9ce06d358ba7","process.pid":67228},"Attributes":{"request":{"status":"Hello"},"task_id":"if","workflow_name":"if_else"}}
. . . . . . . . . . .
{"Body":"World {\"key1\":\"v1\",\"key2\":\"v2\"}","Timestamp":"2024-04-10T09:40:45.127Z000000","SeverityNumber":17,"SeverityText":"ERROR","TraceId":"afde0bf5bb3533d932c1c04c30d91172","SpanId":"ad477b2cf81ca711","TraceFlags":"01","Resource":{"service.name":"unknown_service:node","host.hostname":"9ce06d358ba7","process.pid":67228},"Attributes":{"customer_name":"Hell!","task_id":"if","workflow_name":"if_else"}}