slo-generator 2.0.0rc3

SLO Generator

SLO Generator

slo-generator is a tool to compute and export Service Level Objectives (SLOs), Error Budgets and Burn Rates, using configurations written in YAML (or JSON) format.

Table of contents

• Description
• Local usage
  • Requirements
  • Installation
  • CLI usage
  • API usage
• Configuration
  • SLO configuration
  • Shared configuration
• More documentation
  • Build an SLO achievements report with BigQuery and DataStudio
  • Deploy the SLO Generator in Cloud Run
  • Deploy the SLO Generator in Kubernetes (Alpha)
  • Deploy the SLO Generator in a CloudBuild pipeline
  • DEPRECATED: Deploy the SLO Generator on Google Cloud Functions (Terraform)
  • Contribute to the SLO Generator

Description

The slo-generator runs backend queries computing Service Level Indicators, compares them with the Service Level Objectives defined and generates a report by computing important metrics:

• Service Level Indicator (SLI) defined as SLI = N_{good_events} / N_{valid_events}
• Error Budget (EB) defined as EB = 1 - SLI
• Error Budget Burn Rate (EBBR) defined as EBBR = EB / EB_target
• ... and more, see the example SLO report.

The Error Budget Burn Rate is often used for alerting on SLOs, as it demonstrates in practice to be more reliable and stable than alerting directly on metrics or on SLI > SLO thresholds.

Local usage

Requirements

• python3.7 and above
• pip3

Installation

slo-generator is a Python library published on PyPI. To install it, run:

pip3 install slo-generator

Notes:

• To install providers, use pip3 install slo-generator[<PROVIDER_1>, <PROVIDER_2>, ... <PROVIDER_n]. For instance:
  • pip3 install slo-generator[cloud_monitoring] installs the Cloud Monitoring backend / exporter.
  • pip3 install slo-generator[prometheus, datadog, dynatrace] install the Prometheus, Datadog and Dynatrace, backends / exporters.
• To install the slo-generator API, run pip3 install slo-generator[api].
• To enable debug logs, set the environment variable DEBUG to 1.
• To enable colorized output (local usage), set the environment variable COLORED_OUTPUT to 1.

CLI usage

To compute an SLO report using the CLI, run:

slo-generator compute -f <SLO_CONFIG_PATH> -c <SHARED_CONFIG_PATH> --export

where:

• <SLO_CONFIG_PATH> is the SLO configuration file or folder path.

• <SHARED_CONFIG_PATH> is the Shared configuration file path.

• --export | -e enables exporting data using the exporters specified in the SLO configuration file.

Use slo-generator compute --help to list all available arguments.

API usage

On top of the CLI, the slo-generator can also be run as an API using the Cloud Functions Framework SDK (Flask):

slo-generator api -c <SHARED_CONFIG_PATH>

where:

• <SHARED_CONFIG_PATH> is the Shared configuration file path or GCS URL.

Once the API is up-and-running, you can HTTP POST SLO configurations to it.

Notes:

• The API responds by default to HTTP requests. An alternative mode is to respond to CloudEvents instead, by setting --signature-type cloudevent.

• Use --target export to run the API in export mode only (former slo-pipeline).

Configuration

The slo-generator requires two configuration files to run, an SLO configuration file, describing your SLO, and the Shared configuration file (common configuration for all SLOs).

SLO configuration

The SLO configuration (JSON or YAML) is following the Kubernetes format and is composed of the following fields:

• api: sre.google.com/v2

• kind: ServiceLevelObjective

• metadata:

  • name: [required] string - Full SLO name (MUST be unique).
  • labels: [optional] map - Metadata labels, for example:
    • slo_name: SLO name (e.g availability, latency128ms, ...).
    • service_name: Monitored service (to group SLOs by service).
    • feature_name: Monitored feature (to group SLOs by feature).

• spec:

  • description: [required] string - Description of this SLO.
  • goal: [required] string - SLO goal (or target) (MUST be between 0 and 1).
  • backend: [required] string - Backend name (MUST exist in SLO Generator Configuration).
  • service_level_indicator: [required] map - SLI configuration. The content of this section is specific to each provider, see docs/providers.
  • error_budget_policy: [optional] string - Error budget policy name (MUST exist in SLO Generator Configuration). If not specified, defaults to default.
  • exporters: [optional] string - List of exporter names (MUST exist in SLO Generator Configuration).

Note: you can use environment variables in your SLO configs by using ${MY_ENV_VAR} syntax to avoid having sensitive data in version control. Environment variables will be replaced automatically at run time.

→ See example SLO configuration.

Shared configuration

The shared configuration (JSON or YAML) configures the slo-generator and acts as a shared config for all SLO configs. It is composed of the following fields:

• backends: [required] map - Data backends configurations. Each backend alias is defined as a key <backend_name>/<suffix>, and a configuration map.

  backends:
    cloud_monitoring/dev:
      project_id: proj-cm-dev-a4b7
    datadog/test:
      app_key: ${APP_SECRET_KEY}
      api_key: ${API_SECRET_KEY}
  

  See specific providers documentation for detailed configuration:

  • cloud_monitoring
  • cloud_service_monitoring
  • prometheus
  • elasticsearch
  • datadog
  • dynatrace
  • <custom>

• exporters: A map of exporters to export results to. Each exporter is defined as a key formatted as <exporter_name>/<suffix>, and a map value detailing the exporter configuration.

  exporters:
    bigquery/dev:
      project_id: proj-bq-dev-a4b7
      dataset_id: my-test-dataset
      table_id: my-test-table
    prometheus/test:
      url: ${PROMETHEUS_URL}
  

  See specific providers documentation for detailed configuration:

  • pubsub to stream SLO reports.
  • bigquery to export SLO reports to BigQuery for historical analysis and DataStudio reporting.
  • cloud_monitoring to export metrics to Cloud Monitoring.
  • prometheus to export metrics to Prometheus.
  • datadog to export metrics to Datadog.
  • dynatrace to export metrics to Dynatrace.
  • <custom> to export SLO data or metrics to a custom destination.

• error_budget_policies: [required] A map of various error budget policies.

  • <NAME>: Name of the error budget policy.
    • steps: List of error budget policy steps, each containing the following fields:
      • window: Rolling time window for this error budget.
      • alerting_burn_rate_threshold: Target burnrate threshold over which alerting is needed.
      • urgent_notification: boolean whether violating this error budget should trigger a page.
      • overburned_consequence_message: message to show when the error budget is above the target.
      • achieved_consequence_message: message to show when the error budget is within the target.

  error_budget_policies:
    default:
      steps:
      - name: 1 hour
        burn_rate_threshold: 9
        alert: true
        message_alert: Page to defend the SLO
        message_ok: Last hour on track
        window: 3600
      - name: 12 hours
        burn_rate_threshold: 3
        alert: true
        message_alert: Page to defend the SLO
        message_ok: Last 12 hours on track
        window: 43200
  

→ See example Shared configuration.

More documentation

To go further with the SLO Generator, you can read:

Build an SLO achievements report with BigQuery and DataStudio

Deploy the SLO Generator in Cloud Run

Deploy the SLO Generator in Kubernetes (Alpha)

Deploy the SLO Generator in a CloudBuild pipeline

DEPRECATED: Deploy the SLO Generator on Google Cloud Functions (Terraform)

Contribute to the SLO Generator