Dynatrace Training Platform

SLO Management

Service-Level Objectives track reliability targets. This module covers SLI patterns, burn rate alerting, and the official SLO templates.

SLI DQL Patterns

Custom SLIs must produce an sli field returning an array of double values.

Service Availability

timeseries {
    total = sum(dt.service.request.count),
    failures = sum(dt.service.request.failure_count)
  }, by: { dt.smartscape.service }
| fieldsAdd entityName = getNodeName(dt.smartscape.service)
| fieldsAdd sli = (((total[] - failures[]) / total[]) * 100)
| fieldsRemove total, failures

Service Error Rate by K8s Cluster

timeseries {
    total = sum(dt.service.request.count),
    errors = sum(dt.service.request.failure_count)
  }, by: { dt.smartscape.service, k8s.cluster.name }
| filter k8s.cluster.name == "production-cluster"
| fieldsAdd errorRate = (errors[] / total[]) * 100
| fieldsAdd sli = 100 - errorRate[]

Response Time Performance

timeseries total = avg(dt.service.request.response_time),
  default: 0, by: { dt.smartscape.service }
| fieldsAdd high = iCollectArray(if(total[] > (1000 * 500), total[]))
| fieldsAdd low = iCollectArray(if(total[] <= (1000 * 500), total[]))
| fieldsAdd highRespTimes = iCollectArray(if(isNull(high[]), 0, else: 1))
| fieldsAdd lowRespTimes = iCollectArray(if(isNull(low[]), 0, else: 1))
| fieldsAdd sli = 100 * (lowRespTimes[] / (lowRespTimes[] + highRespTimes[]))

Error Budget & Burn Rate

Concept          What It Means
───────────────  ──────────────────────────────────────────────
Error budget     Amount of "allowed" downtime before breaching target
Burn rate 1.0    Consuming budget exactly at target pace
Burn rate > 1.0  Will breach SLO before evaluation period ends
Burn rate < 1.0  Under budget — room to spare

Dynatrace auto-calculates burn rate and raises events via Anomaly Detection. You can also create custom burn-rate alerts.

Official SLO Templates (7)

Template                              What It Measures
────────────────────────────────────  ──────────────────────────
Host CPU utilization                  CPU idle percentage
Service availability                  Request success rate
Service performance                   Response time threshold
K8s cluster CPU efficiency            Cluster CPU utilization
K8s cluster memory efficiency         Cluster memory utilization
K8s namespace CPU efficiency          Namespace CPU utilization
K8s namespace memory efficiency       Namespace memory utilization

Key Rules

Use timeseries for metric-based SLOs (pre-aggregated, faster, cheaper)
Use makeTimeseries for event/log/span-based SLOs
getNodeName() extracts entity display name from entity ID
SLO names must be unique (duplicate → 400 error)