Fluent Bit scheduling considerations

The default DaemonSet reads logs from the node where it runs. Restricting eligible nodes leaves other nodes’ logs uncollected; adding backend replicas does not fix missing collector coverage.

Scheduling is configured through the package’s upstream passthrough parent. In canonical Big Bang configuration, that parent is packages.fluentbit.values.upstream with packageConfiguration.version: v1; the legacy parent is fluentbit.values.upstream.

Use the bundled chart’s scheduling options and Kubernetes node assignment guidance rather than copying a separate schema here. Check that selectors, affinity, tolerations and node admission permit collection on every intended node.

Return to the integration overview.

For an intentionally limited collector pool, these package values show the scheduling parent without changing privileges or log mounts:

upstream:
  nodeSelector:
    logging.example.com/collect: "true"

Provision this label on every intended node first. Nodes without it will not receive this DaemonSet’s collector; use this only when that coverage boundary is intentional. Existing affinity and taints still apply. This is not a fix for missing host paths, SELinux denials, or insufficient resources.