Skip to content

Slo-exporter computes standardized SLI and SLO metrics based on events coming from various data sources.

License

Notifications You must be signed in to change notification settings

freenowtech/slo-exporter

 
 

Repository files navigation

SLO exporter

CircleCI GitHub release (latest SemVer) Docker Pulls GitHub All Releases

Tool slo-exporter computes standardized Service Level Indicator (SLI) and Service Level Objectives (SLO) metrics based on events coming from various data sources. It follows principles from the SRE Workbook. With already prepared examples, Grafana dashboards, Prometheus recording rules and alerts, you can easily start to alert on SLO in your infrastructure.

If you want to start with computing SLOs, make sure to check out this guide!

Motivation

After more than year of experience of maintaining SLO alerting based on application metrics just from Prometheus, number of issues showed up which made it very difficult and unbearable. Few among others:

  • High cardinality of metrics if we wanted to easily find out which event type caused the alert or affected the error budget.
  • Classification of events ending up as huge regular expressions in the PromQLs.
  • Issues with default values for the computation if no events happened.
  • Need to filter out some events based on high cardinality metadata which cannot be added to metrics.

This lead us to decision that we need to process the events separately and in Prometheus do just the final computation and alerting.

We describe our journey towards SLO based alerting more in detail in the two articles:

How it works

Every ingested event has metadata which are used to classify it to specific SLO domain and class as described in the SRE Workbook chapter Alerting on SLOs. Additionally, name of the application where the event happened and identifier of the event is also added to ease the debugging of possible SLO violation. Finally, you decide based on the metadata if the event was a successful or failed one. Slo-exporter then exposes Prometheus metric slo_domain_slo_class:slo_events_total{slo_domain="...", slo_class="...", result="..."}. This gives you number of successful or failed events which is all you need to calculate the error budget, burn rate etc.

Installing

Build

In the root of the repository run

make

Docker

Prebuilt docker images can be found at Docker Hub.

docker run -it seznam/slo-exporter:<version> --help

Prebuilt binaries

See the the latest release page for the prebuilt binaries.

Configuration and examples

Detailed configuration documentation you can find here docs/configuration.

To see some real use-cases and examples you can look at the examples/.

Operating

Some advices on operating the slo-exporter, debugging and profiling can be found here docs/operating.md.

Please note that features marked as Experimental are not considered stable and may be removed or changed even in minor or patch release.

Community

About

Slo-exporter computes standardized SLI and SLO metrics based on events coming from various data sources.

Resources

License

Code of conduct

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Go 98.0%
  • Shell 1.2%
  • Other 0.8%