Deployment models
The two ways to run nano, Open and Self-run, and where to run it: Docker Compose on one server or several, Kubernetes with Helm, or air-gapped.
Every nano deployment runs the same architecture: Vector receives and parses logs, ClickHouse stores events, and PostgreSQL holds rules, alerts, cases, users, settings and audit metadata. Events are normalized to OCSF and queried with nPL, so searches, detections and dashboards work the same way wherever nano runs.
One deployment serves one organization. Its data, credentials, detections, risk, prevalence and enrichment results are never shared with another deployment.
Two ways to run nano
| Price | Who runs it | Install guide | |
|---|---|---|---|
| Open | Free (AGPL-3.0) | You, on your own servers | Install with Docker Compose |
| Self-run | Paid | You, on your own servers | Start at Self-run, then Docker Compose, High availability or Kubernetes (Helm) |
Both run on servers you control. Pages in these docs marked Self-hosted apply to both; where a feature needs the paid license, the page says so.
What each includes
| Open | Self-run | |
|---|---|---|
| License | AGPL-3.0 | Commercial license |
| Images | ghcr.io/nano-rs/nano-* (public) | nano's licensed registry; credentials tied to your license |
| Ingest, parsing, search, detections, alerting, dashboards | ✓ | ✓ |
| Cases, notebooks, pivt AI, risk analytics, SSO (OIDC), playbooks, custom enrichments | No | ✓ |
| High availability | No | ✓ |
| Air-gapped operation | No | ✓ |
| Runs on | Docker Compose, one server | Docker Compose (one server or several), Kubernetes (Helm) |
Every Self-run deployment runs one build: the same set of images; only the deployment around it differs. Contact hello@nano.rs about commercial licensing.
Deployment options
| Docker Compose | Kubernetes (Helm) | |
|---|---|---|
| Edition | Open or Self-run | Self-run |
| Where it runs | One server, or three or more for Self-run high availability | Your Kubernetes cluster |
| Scale | Up to about 100 GB/day | From one node to multi-shard ClickHouse (10 TB/day+) |
| High availability | Self-run across three or more servers: ClickHouse replicas, Keeper quorum, several ingest servers, and from v2.0.7 several app servers and PostgreSQL failover. See High availability | Optional ClickHouse replicas, multiple app replicas, external PostgreSQL |
| Ingest | One server: Vector ports or /ingest through nginx. Several servers (Self-run): a dedicated ingest frontend | A dedicated ingest frontend with its own hostname |
| Guide | Install with Docker Compose | Install on Kubernetes (Helm) |
Docker Compose
The installer sets up the whole stack on one server: nginx, the API, search, jobs and web services, ClickHouse, PostgreSQL, Vector and Valkey. Open installs from the public nano-rs/nano repository; Self-run installs from the Self-run installer bundle with your license values. Both verify every image against a committed digest lock before starting.
curl -fsSL https://get.nano.rs | bashSee Install with Docker Compose for requirements, sizing and production setup. For Self-run across three or more servers, see High availability.
Kubernetes with Helm (sovereign)
The nano Helm chart runs Self-run on your own cluster. ClickHouse and Keeper run through the official ClickHouse operator, PostgreSQL is bundled or external, and the chart ships sizing presets from about 10 GB/day to about 10 TB/day. Ingest gets its own hostname, Ingress or load balancer, edge pods and node pool, so a flood of events can't take the app down. See Install on Kubernetes (Helm).
Air-gapped
Self-run also runs in disconnected networks, with a signed image tarball, signed offline content bundles, an offline license and on-premises AI. See Air-gapped deployment.
Storage and query
- Events live in the ClickHouse table
nanosiem.ocsf_logs, partitioned by day. Hot OCSF paths (endpoints, users, hosts, processes, files and more) are stored as typed, indexed columns; the full event is kept as JSON, and fields the parser didn't map stay underunmapped. See OCSF schema. - One table holds every OCSF version: each row keeps its own
metadata.version, so a new OCSF release needs no migration of stored data. - nPL compiles to ClickHouse SQL. Searches always carry a time range, so ClickHouse reads only the partitions in it.
- Detection findings, platform audit, prevalence and identity data are also stored in ClickHouse and searchable with nPL.
- OTLP traces and metrics go to their own tables and power the
services,service,traceandmetricsearch commands.
Ingest
Vector receives and parses events on every deployment.
| Protocol | Docker Compose, one server | Kubernetes, or Compose with several servers | Auth |
|---|---|---|---|
| HTTP (JSON / NDJSON) | http(s)://<host>/ingest, or :8080 | https://<ingest-host>/ingest | Authorization: Bearer <token> |
| HEC | :8088/services/collector/event | https://<ingest-host>/services/collector | Authorization: Splunk <token> |
| OTLP/HTTP | http(s)://<host>/otlp/v1/{logs,traces,metrics} | https://<ingest-host>/otlp/v1/{logs,traces,metrics} | Source-bound ingest token |
| Raw OTLP, Fluent Forward | 127.0.0.1 by default; opt in to other interfaces | Cluster-internal unless you expose them | None |
Vector native (:6000) | Every interface (from v2.0.30) | On each ingest server (multi-server); in-cluster on Kubernetes when enabled | Mutual TLS: a client certificate per sender |
| Pull sources | Kafka, AWS S3/SQS, GCP Pub/Sub, SaaS pull feeds | Same | Per-source credentials |
Ingest tokens are bound to one log source each and issued from Ingestion → Log Sources → Credentials. Unauthenticated listeners (raw OTLP, Fluent Forward) are for trusted networks only. Vector native requires a client certificate from v2.0.30; see Vector native. Syslog ships disabled; enable it in Vector, or send syslog to a collector that forwards over HTTP or HEC. See Integrations.
Retention
ClickHouse keeps events for 365 days by default and deletes them by daily partition when they age out. You can change the window, move older data to S3-compatible object storage, and let disk pressure protect the local disk. See Storage & Retention.
Choose a deployment
- Docker Compose to try nano, run Open, or run Self-run on one server up to about 100 GB/day, or on three or more for high availability.
- Kubernetes (Self-run) if you already run clusters, or ingest more than one server can handle.
- Air-gapped (Self-run) if the deployment can't reach the internet.
What's next
Add a SIEM to Your Ludus Cyber Range
Send logs from your Ludus cyber range into nano SIEM, then search them and run detections against the attacks you launch in the lab.
OCSF schema
How nano models events with OCSF 1.9 — classes, log sources, class-dependent paths, unmapped fields, raw data, enrichments and fast filters.