Standalone vs. Integrated Deployment Modes
The charts/nxtflow Helm chart supports three deployment topologies controlled by tags.standalone and externalLoki.url.
Mode Comparison
| Feature | Standalone (tags.standalone=true) |
Integrated (tags.standalone=false) |
Collector-Only (tags.standalone=false) |
|---|---|---|---|
goflow2 + nxtflow-enricher |
Deployed | Deployed | Deployed |
| Bundled Loki (SingleBinary) | Deployed | Disabled | Disabled |
| Bundled Prometheus | Deployed (server only) | Disabled | Disabled |
| Bundled Grafana + Dashboard | Deployed (volkovlabs-echarts-panel) |
Disabled | Disabled |
ServiceMonitor & PrometheusRule |
Optional | Enabled (monitoring.*.enabled=true) |
Disabled |
| Grafana Dashboard ConfigMap | Auto-provisioned | Enabled (grafana.dashboardConfigMap.enabled=true) |
Disabled |
nxtflow-mcp Server |
Enabled | Enabled | Disabled |
| Example Values File | examples/values-standalone.yaml |
examples/values-integrated.yaml |
examples/values-collector-only.yaml |
0. Docker Compose (Single-Host / Homelab)
For single-host Docker setups without Kubernetes, use the root docker-compose.yml and .env.example — see the full Docker Compose Deployment Guide for the complete .env reference, router switching, and troubleshooting:
cp .env.example .env
./scripts/up.sh --demo
./scripts/verify-stack.sh
All host ports (NXTFLOW_GRAFANA_PORT, NXTFLOW_LOKI_PORT, NXTFLOW_NETFLOW_PORT, NXTFLOW_SFLOW_PORT, NXTFLOW_PROMETHEUS_PORT, NXTFLOW_MCP_PORT), NXTFLOW_SITE, NXTFLOW_LOKI_RETENTION, and Grafana admin credentials are configured in .env. Grafana automatically provisions the Loki and Prometheus datasources alongside nxtflow-overview.json.
1. Standalone Mode
Standalone mode is ideal for personal labs, test clusters, or dedicated monitoring namespaces without an existing observability stack:
helm install nxtflow oci://ghcr.io/nertwork/charts/nxtflow \
--namespace nxtflow --create-namespace \
-f examples/values-standalone.yaml
Bundled subcharts (grafana-community/loki, prometheus-community/prometheus, grafana-community/grafana) are automatically wired via configmap-grafana-datasources.yaml using your release name (http://<release>-loki:3100 and http://<release>-prometheus-server:80).
2. Integrated Mode
Use Integrated mode when your cluster already runs Loki, kube-prometheus-stack, and Grafana with sidecar discovery:
helm install nxtflow oci://ghcr.io/nertwork/charts/nxtflow \
--namespace monitoring \
-f examples/values-integrated.yaml \
--set externalLoki.url="https://loki.example.com"
Configuration Guardrails
The chart enforces fail-fast checks in _helpers.tpl:
- Setting both tags.standalone=true and externalLoki.url fails immediately to prevent split-brain routing.
- Setting tags.standalone=false without externalLoki.url fails unless enricher.shipper.mode="file".
- Enabling mcp.enabled=true without either bundled Loki or externalLoki.url fails immediately.
3. Collector-Only Mode (Multi-Site Edge)
Deploy lightweight collectors at remote clusters or branch sites shipping into a central Loki instance:
helm install nxtflow-branch oci://ghcr.io/nertwork/charts/nxtflow \
--namespace nxtflow \
-f examples/values-collector-only.yaml \
--set collector.site="branch-west" \
--set externalLoki.url="https://loki.example.com"
Every site registers dynamically in the Grafana Site dropdown via label_values({job="nxtflow"}, site) and appears in nxtflow status.
Sankey Visualization Engine (echarts vs. netsage)
echarts(default): Uses the signedvolkovlabs-echarts-panelplugin with bipartite node IDs (s:<src>->d:<dst>) so bidirectional traffic never triggers ECharts DAG cycle errors.netsage(opt-in): Setgrafana.sankey.engine=netsageto render thenetsage-sankey-panelpanel JSON (requiresnetsage-sankey-panelinstalled and allowed underGF_PLUGINS_ALLOW_LOADING_UNSIGNED_PLUGINSin your Grafana deployment).
Alloy Sidecar Mode
Set enricher.shipper.mode=alloy if you prefer nxtflow-enricher to write JSONL files to /var/log/nxtflow/flows.jsonl and delegate Loki shipping to a grafana/alloy sidecar container.