Node Agent Configuration

The node agent reads its configuration from environment variables, which you set through nodeAgent.env in your values file.

Some of these have a dedicated chart value and are easier to set that way: nodeAgent.scrapeInterval, nodeAgent.apiKey, nodeAgent.tracesEndpoint, nodeAgent.logsEndpoint, nodeAgent.metricsEndpoint, nodeAgent.profilesEndpoint. The chart also passes --cgroupfs-root /host/sys/fs/cgroup as a command-line flag, which wins over CGROUPFS_ROOT, so leave that one alone.

Network and listening

Flag	Description	Default
`LISTEN`	Address the agent listens on (`ip:port` or `:port`)	`0.0.0.0:80`
`CGROUPFS_ROOT`	Path to cgroup filesystem mount	`/sys/fs/cgroup`
`EPHEMERAL_PORT_RANGE`	Skip tracking these TCP port ranges	`32768-60999`
`TRACK_PUBLIC_NETWORK`	Whitelisted public IP networks to track	`0.0.0.0/0`

Metrics and logs

Flag	Description	Default
`DISABLE_LOG_PARSING`	Disable container log parsing	`false`
`DISABLE_L7_TRACING`	Disable Layer 7 (application-level) tracing	`false`
`DISABLE_PINGER`	Disable upstream pinging	`true`
`LOG_PER_SECOND`	Max number of logs processed per second	`10.0`
`LOG_BURST`	Max log burst capacity (tokens)	`100`
`MAX_LABEL_LENGTH`	Maximum length for any metric label value	`4096`
`EXCLUDE_HTTP_REQUESTS_BY_PATH`	Skip HTTP metrics by path (comma-separated)	(none)

Redaction

Flag	Description	Default
`SANITIZE_HEADERS`	Enable header sanitization	`true`
`SENSITIVE_HEADERS`	List of sensitive header names to sanitize (comma-separated)	`Authorization, Cookie, X-Action-Token`
`DISABLE_SENSITIVE_LOG_PARSING`	Stop parsing log lines that look like they carry secrets	`true`

Cloud labels

These enrich the node_cloud_info metric:

Flag	Description
`PROVIDER`	Cloud provider name
`REGION`	Cloud region
`AVAILABILITY_ZONE`	Availability zone
`ACCOUNT_ID`	Cloud account ID
`INSTANCE_TYPE`	Instance type
`INSTANCE_LIFE_CYCLE`	Instance lifecycle (spot/on-demand)

Endpoints

Flag	Description
`COLLECTOR_ENDPOINT`	Base endpoint for all telemetry
`API_KEY`	API key for the endpoints above. Set it with `nodeAgent.apiKey`.
`METRICS_ENDPOINT`	Specific metrics push endpoint
`TRACES_ENDPOINT`	Specific traces push endpoint
`LOGS_ENDPOINT`	Specific logs push endpoint
`PROFILES_ENDPOINT`	Specific profiles push endpoint
`INSECURE_SKIP_VERIFY`	Skip TLS verification (not recommended)

Scraping and storage

Flag	Description	Default
`SCRAPE_INTERVAL`	How often to gather metrics	`15s`
`WAL_DIR`	Where to store Write-Ahead Logs	`/tmp/coroot-node-agent`

Other flags

Flag	Description	Default
`RESOLVE_DNS`	Enable DNS resolution	`false`
`IGNORE_CONTROL_PLANE`	Ignore these control plane components	`karpenter,loki,prometheus,grafana,kubelet,etcd,apiserver,victoria,nudgebee-agent,kube-system`
`TRACE_ID_HEADERS`	Headers to extract trace IDs	`Traceparent,X-Request-Id`

Applying configuration

Here’s an example values.yaml with common flags explicitly set:

Do not change LISTEN

The DaemonSet declares container port 80 and the PodMonitor scrapes the port named http, which is that one. Moving the listener to another port leaves Prometheus scraping a port nothing is bound to, and network metrics go quiet with no error.

nodeAgent:
  env:
    - name: DISABLE_LOG_PARSING
      value: "true"
    - name: LOG_PER_SECOND
      value: "20"
    - name: SANITIZE_HEADERS
      value: "true"
    - name: SENSITIVE_HEADERS
      value: "Authorization,Proxy-Authorization,Cookie,X-Auth-Token"
    - name: PROVIDER
      value: "aws"
    - name: REGION
      value: "us-east-1"
    - name: ACCOUNT_ID
      value: "123456789"
    - name: SCRAPE_INTERVAL
      value: "30s"

To apply these settings, run:

helm upgrade --install nudgebee-agent nudgebee-agent/nudgebee-agent \
  --namespace nudgebee-agent --create-namespace \
  -f values.yaml

Network and listening​

Metrics and logs​

Redaction​

Cloud labels​

Endpoints​

Scraping and storage​

Other flags​

Applying configuration​