Loki¶

Loki consumes shared networking and Istio resources through the bb-common subchart. Configure these with bb-common.istio, bb-common.networkPolicies, and bb-common.routes; see the migration guide when upgrading from the former root-level values.

Overview¶

This package contains an extensible and configurable abstract installation of Grafana Loki based on the upstream chart provided by grafana.

For Loki architecture, sizing, queries, LogQL, and the full configuration reference, use the upstream Loki documentation.

Big Bang integration¶

Big Bang deploys Loki as the logging-loki release in the logging namespace. The umbrella chart owns package enablement, source selection, release ordering, and values assembly. Configure the integration through the canonical packages.loki value; the legacy loki value remains supported for compatibility.

The selected package version is a patched upstream chart, not an upstream passthrough wrapper. For the complete chart configuration surface, use the versioned upstream Loki Helm values. Big Bang-specific package values and overrides are defined in this repository’s chart/values.yaml.

Storage and deployment strategy¶

Big Bang supports monolith and scalable strategies. Monolith uses filesystem storage only when the embedded MinIO instance is disabled; otherwise, and for scalable deployments, Loki requires S3-compatible object storage. The umbrella maps packages.loki.objectStorage endpoint, region, credentials, and bucket names into the chart. Omit static credentials only when the selected object-store identity mechanism supplies them. Treat schema, bucket, retention, and compactor settings as data-safety interfaces and plan upgrades using the upstream upgrade guidance.

distributed can render but is unsupported by Big Bang. It is not a validated production strategy.

For example, these umbrella values select the supported scalable strategy with existing external storage. Strategy and storage inputs are siblings of values, not chart options under an upstream parent:

packageConfiguration:
  version: v1
packages:
  loki:
    enabled: true
    strategy: scalable
    objectStorage:
      endpoint: https://objects.example.com
      region: us-east-1
      bucketNames:
        chunks: loki-chunks
        ruler: loki-ruler
        admin: loki-admin
    values:
      minio:
        enabled: false
      backend:
        replicas: 3

Provision the actual buckets and permissions separately. Replace the endpoint, region, and bucket names; supply static credentials only through protected objectStorage.accessKey/accessSecret inputs, or configure and verify the environment’s workload identity. An omitted credential is not proof that identity works. Keep backend replicas/autoscaling minima within the package production guard; do not use experimentalMode to bypass it for production.

For a separate filesystem-backed monolith installation, the umbrella values shape is:

packageConfiguration:
  version: v1
packages:
  loki:
    enabled: true
    strategy: monolith
    values:
      minio:
        enabled: false
      singleBinary:
        persistence:
          enabled: true
          size: 40Gi

The capacity is illustrative, not sizing guidance. Filesystem storage is not a redundant object store. These alternatives are not a safe in-place strategy or storage migration procedure; preserve data, schema history, collector endpoints, network access, and recovery requirements when planning a change. The umbrella maps the strategy to chart deploymentMode and component replicas; direct chart users must configure those together rather than setting a chart strategy key.

Loki retention is enabled through the compactor by default, while the retention period remains an operator decision. Review the upstream retention and compactor documentation before changing retention or deletion behavior.

Ingestion, tenancy, and access¶

Loki does not collect logs. Big Bang can route logs from Alloy or Fluent Bit to Loki; their collection configuration and lifecycle remain owned by their respective packages. Grafana is the query interface and datasource consumer; Grafana authentication and SSO are owned by the monitoring and Grafana packages, not by Loki.

The default Loki configuration disables authentication. Enabling Loki or Grafana Enterprise Logs tenancy/authentication requires a complete, package-specific access model and any required license material; use the upstream authentication configuration and Grafana Enterprise Logs documentation for those product settings. Do not treat Loki authentication as a replacement for Grafana or mesh access controls.

Networking, mesh, and observability¶

Big Bang creates the Loki route, network-policy inputs, dashboards, and monitoring resources from package values. The route targets the monolith service for the monolith strategy and the gateway for scalable or distributed rendering. Network access is limited for configured collectors, Grafana, Prometheus, MinIO, Tempo, and Big Bang tests; extend it deliberately when adding other clients or external storage networks.

When Istio is enabled, Big Bang supplies the chart’s Istio integration and mTLS-aware monitoring configuration. Mesh and gateway policy ownership remains in the umbrella and Istio packages. Loki exposes readiness at /ready; the chart also creates a ServiceMonitor when metric scraping is available. Dashboards and datasource checks integrate with Grafana, but neither proves log ingestion or object-store durability.

Operational boundaries¶

Use upstream Loki documentation for architecture, sizing, queries, LogQL, configuration reference, and operational runbooks. Big Bang validates its supported monolith and scalable package lanes, but production validation should include the chosen collector, object storage, Grafana queries, retention behavior, and network policy/mesh path. Tempo, ECK, Alloy, and Fluent Bit have separate package ownership and are not configured by this guide.