Agent (DaemonSet — collect logs from all nodes)

helm install vector-agent vector/vector
--namespace vector --create-namespace
--set role=Agent
--set customConfig.sources.kubernetes_logs.type=kubernetes_logs
--set customConfig.sinks.stdout.type=console
--set 'customConfig.sinks.stdout.inputs[0]=kubernetes_logs'
--set customConfig.sinks.stdout.encoding.codec=json


## Roles

| Role | Workload | Use Case |
|------|----------|----------|
| `Agent` | DaemonSet | Collect logs/metrics from every node |
| `Aggregator` | StatefulSet | Centralized processing with persistent buffer |
| `Stateless-Aggregator` | Deployment | Centralized processing without persistence |

```bash
# Aggregator (StatefulSet — centralized processing)
helm install vector-aggregator vector/vector \
  --namespace vector \
  --set role=Aggregator \
  --set replicas=3 \
  --set customConfig.sources.vector.type=vector \
  --set customConfig.sources.vector.address=0.0.0.0:6000 \
  --set customConfig.sinks.stdout.type=console \
  --set 'customConfig.sinks.stdout.inputs[0]=vector' \
  --set customConfig.sinks.stdout.encoding.codec=json

Configuration

Pipeline via customConfig

customConfig:
  data_dir: /vector-data-dir
  api:
    enabled: true
    address: 127.0.0.1:8686
  sources:
    kubernetes_logs:
      type: kubernetes_logs
      extra_label_selector: "app!=vector"
    internal_metrics:
      type: internal_metrics
    vector:
      address: 0.0.0.0:6000
      type: vector
      version: "2"

  transforms:
    remap:
      type: remap
      inputs: [kubernetes_logs]
      source: |
        .@timestamp = del(.timestamp)
        .service = .kubernetes.container_name

  sinks:
    console:
      type: console
      inputs: [remap]
      encoding:
        codec: json
    prometheus:
      type: prometheus_exporter
      inputs: [internal_metrics]
      address: 0.0.0.0:9090

HAProxy Load Balancer (Aggregator)

haproxy:
  enabled: true
  replicas: 2
  service:
    type: ClusterIP

PodMonitor (Prometheus)

podMonitor:
  enabled: true
  interval: 30s
  path: /metrics
  port: prom-exporter

Key Parameters

Parameter	Default	Role	Description
`role`	`Aggregator`	all	Workload type
`image.repository`	`docker.io/timberio/vector`	all	Image repo
`image.tag`	derived from appVersion	all	Image tag
`image.base`	`distroless-libc`	all	Base distro
`replicas`	`1`	Agg/Stateless	Pod count
`customConfig`	`{}`	all	Pipeline definition
`existingConfigMaps`	`[]`	all	Use external ConfigMap for config
`dataDir`	`""`	all	Data dir (required with existingConfigMaps)
`secrets.generic`	`{}`	all	Key-value secrets
`persistence.enabled`	`false`	Aggregator	PVC for buffer
`persistence.size`	`10Gi`	Aggregator	PVC size
`service.enabled`	`true`	all	Create Service
`service.type`	`ClusterIP`	all	Service type
`rbac.create`	`true`	Agent	RBAC for pod/node access
`psp.create`	`false`	Agent	PodSecurityPolicy
`autoscaling.enabled`	`false`	Agg/Stateless	HPA
`podMonitor.enabled`	`false`	all	Prometheus PodMonitor
`haproxy.enabled`	`false`	Aggregator	HAProxy sidecar
`podLabels`	`{vector.dev/exclude: "true"}`	all	Labels on pods
`env`	`[]`	all	Environment variables
`envFrom`	`[]`	all	Env from Secrets/ConfigMaps
`extraVolumes`	`[]`	all	Additional volumes
`extraVolumeMounts`	`[]`	all	Additional mounts
`initContainers`	`[]`	all	Init containers
`extraContainers`	`[]`	all	Sidecar containers
`resources`	`{}`	all	CPU/memory requests/limits
`nodeSelector`	`{}`	all	Node selection
`tolerations`	`[]`	all	Taints
`affinity`	`{}`	all	Scheduling affinity
`updateStrategy`	`{}`	all	Rolling update strategy
`terminationGracePeriodSeconds`	`60`	all	Graceful shutdown
`livenessProbe`	`{}`	all	Liveness
`readinessProbe`	`{}`	all	Readiness
`startupProbe`	`{}`	all	Startup
`extraObjects`	`[]`	all	Extra K8s manifests

Secrets

secrets:
  generic:
    datadog_api_key: "your-api-key"
    aws_access_key_id: "AKIA..."
    aws_secret_access_key: "..."

Referenced via env:

env:
  - name: DATADOG_API_KEY
    valueFrom:
      secretKeyRef:
        name: vector
        key: datadog_api_key

Upgrading

helm repo update
helm upgrade vector-agent vector/vector \
  --namespace vector \
  --values values.yaml \
  --version 0.56.0

Common Mistakes

Role not set — Default is Aggregator. Set --set role=Agent for DaemonSet.
customConfig replaces all defaults — When using customConfig, you must define the full pipeline (sources, transforms, sinks). No defaults are merged.
Agent requires RBAC — rbac.create=true (default) for reading pod logs. Disable only if using pre-created RBAC.
API health endpoint — customConfig.api.enabled=true required for liveness/readiness probes. Without it, probes fail.
gRPC probes — Vector 0.55+ supports gRPC health probes via grpc.health.v1.Health/Check on port 8686.
Aggregator without persistence — persistence.enabled=false means buffer data is lost on pod restart.
HostPath for Agent — Agent uses /var/log/, /var/lib/, /proc, /sys mounts by default. Don't override unless you know the impact.
Base distribution — distroless-libc is minimal. Use debian if you need shell access or debugging tools.

vector-helm

Vector — Helm Chart

Quick Install