Connect SLA targets to an incident response process
An availability figure needs a clear measurement scope. An application outage also needs observation, ownership and a plan to resume operation.
An availability figure is easier to compare than the organisation of an incident response. When running your application, however, you need to know what the figure measures and how a noteworthy observation leads to coordinated action.
Google SRE distinguishes a Service Level Indicator (SLI) as a measure, a Service Level Objective (SLO) as a target and a Service Level Agreement (SLA) as a service agreement. A metric needs a defined scope and measurement window. The term SLA alone does not explain which part of your application workflow is observed.
An available component can conceal an interrupted workflow
In this fictional example, a portal uses different services for its web application and file storage. During event S-60, the homepage continues to respond. New requests with files, however, cannot be saved completely. A measurement point covering only the homepage does not reflect this impact.
The team should therefore relate measurement to the user function it needs. This makes no claim about a particular provider agreement: the example concerns which technical observation is needed for the team's own application.
Add task ownership to the metric
The following plan connects the measurement scope with the response to this case.
Aspect | What to clarify for the portal | Responsible role |
|---|---|---|
Observation | Is only a component checked, or the complete request workflow? | Application owner |
Response | Who brings findings together, contacts the affected services and coordinates changes? | Incident coordination |
Resumption | Who checks the request, file and outstanding work, and decides the next operational step? | Business and operations owners |
These are suggestions for the example, rather than service values measured here or roles already assigned. An availability metric alone determines neither the technical action nor the specific time of resumption.
Review measurement and procedure together
Organising your own incident response does not make external service commitments meaningless. It adds work on the complete workflow to the view of the affected components. Clarify measurement, contact routes and decision-making authority together.
When several services are involved, your team needs a coordinated view of the impact. A technical finding becomes a sound operational decision when it relates to the affected workflow and the next action is assigned to a responsible role.
Sources
Read more
All articles
Cloud & Operations with ASES
Hosting, deployment, maintenance and support