Datadog P95 Latency Check

Last updated on Jul 16, 2026

---

Datadog APM P95 latency check validates the 95th percentile latency of a service using Datadog APM metrics. The target service and environment are supplied at runtime via probe variables, allowing this probe to be reused across services.

Infrastructure type

• Kubernetes

Use cases

Datadog P95 Latency Check probe helps you:

• Verify tail latency stays within SLO during network chaos
• Monitor p95 latency after dependency failures
• Validate latency recovery after fault injection
• Detect latency regressions during rollout chaos

---

Overview

This probe queries Datadog APM for the latency_p95 stat of the target service, converts seconds to milliseconds using the formula p95*1000, and compares the mean aggregated value against your threshold.

Probe type

APM Probe (Datadog)

Prerequisites

• Kubernetes chaos infrastructure with Harness Delegate
• A configured Datadog connector
• Datadog APM instrumented on the target service
• Probe variable SERVICE_NAME set at experiment runtime

---

Probe properties

Comparator

Type	Criteria	Default value
float	<=	500

The probe passes when the mean p95 latency (in milliseconds) is less than or equal to the comparator value.

Datadog APM probe inputs

Field	Description	Default
connectorID	Identifier of the Datadog connector	Required at runtime
durationInMin	Look-back window in minutes	10
queryType	Datadog query API version	v2
queries[0].name	Query alias used in the formula	p95
queries[0].dataSource	Metric source type	apm_metrics
queries[0].params.env	Datadog APM environment	<+probe.variables.ENV>
queries[0].params.service	Datadog APM service name	<+probe.variables.SERVICE_NAME>
queries[0].params.stat	APM latency stat	latency_p95
formula	Converts seconds to milliseconds	p95*1000
aggregation	Aggregation over the time window	mean

Probe variables

Variable	Description	Required
SERVICE_NAME	Datadog APM service name	Yes
ENV	Datadog APM environment tag	No

Configurable inputs

Input	Description	Required	Default
CONNECTOR_ID	Datadog connector identifier	Yes	-
DURATION_IN_MIN	Look-back window in minutes	No	10
COMPARATOR_VALUE	p95 latency threshold in milliseconds	Yes	500

---

Run properties

Property	Description	Type	Default
timeout	Maximum time to wait for the probe to complete	String	30s
interval	Time between probe executions	String	5s
attempt	Number of retry attempts before marking as failed	Integer	1
pollingInterval	Time between retry attempts	String	30s
initialDelay	Initial delay before starting the probe	String	1s
stopOnFailure	Stop the experiment if the probe fails	Boolean	false
verbosity	Log verbosity level	String	debug

---

Probe definition

You can define this probe in your chaos experiment as follows:

probe:
  - name: datadog-p95-latency-check
    type: apmProbe
    mode: Continuous
    apmProbe/inputs:
      type: Datadog
      comparator:
        type: float
        value: "500"
        criteria: <=
      datadogApmProbeInputs:
        connectorID: <+connector.identifier>
        durationInMin: 10
        queryType: v2
        queries:
          - name: p95
            dataSource: apm_metrics
            params:
              env: <+probe.variables.ENV>
              service: <+probe.variables.SERVICE_NAME>
              stat: latency_p95
        formula: p95*1000
        aggregation: mean
    runProperties:
      timeout: 30s
      interval: 5s
      attempt: 1
      pollingInterval: 30s
      initialDelay: 1s
      stopOnFailure: false
      verbosity: debug

---

Next steps

• Go to Datadog P99 Latency Check to validate p99 latency.
• Go to Datadog APM Probe Templates to browse all Datadog probe templates.