The following methods are available to configure APM sampling:

• Single Rule properties: An easy way to define one rule applicable to all traces.
• Configuration File: A comprehensive way to define samplings including branching rules and operation priority rules.

There are 2 sampling strategies:

• Sampling: It uses the sampling entry and it defines a rule tree for trace evaluation.
• Operation Priority: It uses the operation_priority entry and when it's defined, it overrides sampling if trace root span operation name matched defined name. Each defined operation name has its own decision rule tree.

• If the service node (service:) is not specified, then the sampling config is applied to all services without dedicated sampling config. When multiple configs have the same service name, then only the last config with that service name is used.

  If a file has configs with and without this service node, the config with the service name matching the APM Agent or APM Tracer service name is prioritized.

• When sampling (sampling:) and operation_priority (operation_priority:) nodes are both specified, the operation_priority is prioritized. If the trace root span name matched the operation name from operation_priority node, it is evaluated with the rule associated with the operation name. If the trace root span name did not match any operation names from the operation_priority node, then the sampling node is evaluated.

• If sampling node (sampling:) is not specified, or no rule is specified, then the trace is not sampled when no operation_priority rule is applicable.

• When the operation_priority node (operation_priority:) is specified, if the operation name of the trace root span matches any operation_priority operation names, then the rule of matching name is used for evaluation instead of the base sampling rule.

• Rule Node
• Source Value

1. Rule that apply to all services

   sampling_config:
     -
        sampling:
          # Every trace has 2% chance to be sampled.
          rule: probabilistic
          param: 0.02

2. Probabilistic with minimum count

   sampling_config:
     -
        sampling:
         # For every 60 seconds duration, 5 root spans of each operation name are sampled,
         # followed by 10% chance of sampling.
         rule: per-operation-rate
         param: 5/60
         true:
         false:
           rule: probabilistic
           param: 0.1

3. Sample only specified operations

   sampling_config:
     -
        service:
           - order_dept
           - shipping_dept
        operation_priority:
          # For /order operation trace root span, 1 of every 10 traces is sampled.
          - name: "/order"
            rule: constant
            param: 10
          # For /ship operation trace root span, 1 of every 50 traces is sampled.
          - name: "/ship"
            rule: constant
            param: 50
          # For /status_order or /status_ship operations, there is a 50% sampling chance.
          - regex: "/status_(order|ship)"
            rule: probabilistic
            param: 0.5

4. Other sample example

   sampling_config:
     -
       # With service node, this config is only applied to service s1 and s2.
       service:
         - s1
         - s2
       sampling:
         # 5 root span of any operation name are sampled every 60 seconds. Starting from
         # the 6th span of an operation name, there is a 20% chance the span is sampled.
         rule: per-operation-rate
         param: 5/60
         false:
           rule: probabilistic
           param: 0.2
       operation_priority:
         # For /order operation, 1 of every 10 spans is sampled.
         - name: "/order"
           rule: constant
           param: 10
         # For /ship operation, first 3 are sampled per 60 seconds and then 10% chance after
         - name: "/ship"
           rule: limited-rate
           param: 3/60
           false:
             rule: probabilistic
             param: 0.01
     -
      # With service node, this config is only applied to service s3 and s4.
      service:
        - s3
        - s4
      sampling:
        # There is a 5% chance any trace is sampled.
        rule: probabilistic
        param: 0.05
     -
      # Without service node, this config is applied to any service except s1, s2, s3 and s4.
      sampling:
        # There is a 10% chance any trace is sampled.
        rule: probabilistic
        param: 0.1

See below an example of a rule tree used for evaluation:

rule: per-operation-rate
param: 5/60
false:
     rule: probabilistic
     param: 0.2