Service level indicators
A service objective is burning. Find the affected journey
Start with the operations behind the signal, then use infrastructure evidence to locate the cause. The objective tells you about impact, not automatically the failing component.
In this article
Confirm the measurement is current
Check the indicator definition, observation window and telemetry freshness. Confirm that the alert refers to the intended environment and workload.
Inspect numerator and denominator counts. A low-volume period or missing source can change how a rate should be interpreted.
Do not dismiss user reports because the dashboard appears healthy. The indicator may have a coverage gap or a delayed observation window.
Identify the affected population
Break down the result by a small set of useful dimensions such as operation type, region or release. Keep customer-level details in controlled diagnostics.
Find representative failed or late operations and trace their paths. A broad service percentage may be driven by one important endpoint or one queue that has stopped progressing.
Compare recent releases, dependency changes and load patterns with the start of degradation.
Limit ongoing impact
Use the service's tested recovery action, such as stopping a bad rollout, reducing nonessential load or restoring a failed dependency. Choose based on the observed failure rather than the indicator alone.
For an illustrative report-generation backlog, adding web instances will not help if workers are blocked on a provider quota. Inspect the constrained stage before increasing capacity.
Preserve operation identities and safe retry behaviour during recovery.
Confirm recovery without erasing history
Run current workflow checks and observe new operations completing within the required bounds. A rolling objective can remain below target after the immediate incident is resolved because it still includes earlier failures.
Report present health and historical objective impact separately. Do not alter the denominator or reset the window simply to make the display green.
Record the user impact, cause and any measurement gap discovered. The incident should improve both the service and the team's ability to recognise the next failure accurately.
Primary sources
Google SRE: alerting on service objectivesReferences checked 11 September 2026.