Pipeline SLAs #107
Description
Allow users to define a time-based SLA on a Pipeline.
Abstract
- Pipelines are multi-stage, and stage handlers may exist as part of various different microservices each with their own independent failure modes, SLAs, deployments models and so on.
- A Pipeline must be treated as a cohesive whole even though it is composed of disparate parts (stages).
- Pipeline SLAs provide a well-defined target for all Pipeline stage handlers to uphold.
- If a Pipeline instance is taking longer to complete than the defined SLA, then Hadron will generate monitoring alerts with parameterizable messaging indicating the problem.
- Teams can use this to hook the Hadron monitoring endpoints into prometheus and the like in order to generate alerts for paging / alerting.
Future
- Pipeline SLAs v2 will allow for SLAs on individual stages.
- Stage SLAs can have their own messaging for alerts. This allows for different teams to be paged, or other more granular actions to be taken.