nano SIEM
Releases

nano v2.0.23

Supported in:Self-hosted

nano v2.0.23: Agent Lane gives AI agents read-only, audited ClickHouse access over their own listener, and Vector no longer drops events when ClickHouse refuses its writes.

v2.0.23 includes everything in v2.0.22.

To upgrade, follow Upgrade a self-hosted deployment. v2.0.20 is a required stop: installs on a release older than v2.0.20 upgrade to v2.0.20 first, then to v2.0.23. scripts/nano upgrade and helm upgrade refuse to skip it.

New

  • Agent Lane (Self-run, off by default). A dedicated, read-only lane for AI agents and scripts to run hunts in native ClickHouse SQL.
    • Agents connect to a separate HTTPS listener on ClickHouse (port 8444 by default). nano's own ClickHouse ports stay internal.
    • The listener stays closed until you turn it on and give an IP allow-list. Allow-lists that admit everyone (any /0) are refused.
    • Each agent gets its own named, expiring token, with its own optional IP allow-list. You can revoke a token at any time.
    • The number of active tokens is capped from the memory available to ClickHouse, so agents can't starve ingest and search.
    • Every query an agent runs is recorded in a per-token query audit.
    • Settings offers an agent kit to download, with connection details and the live schema.
    • The setup guide ships in the release bundle at docs/administration/agent-lane.md. On a single host, if your agents are on a 192.168.x or 172.16-31.x network, move Docker's address pools first, as the guide describes.

Fixes

  • Vector dropped events when ClickHouse refused its writes. If ClickHouse refused Vector's login or writes, for example because of a network allow-list or password change, Vector discarded batches it had already acknowledged to senders. Vector now holds and retries them until ClickHouse accepts them. In testing, all events sent during a refusal were stored exactly once after access returned. Vector is now 0.61.0-nano.12.
  • Compose upgrades kept serving old nginx config. After an upgrade, nginx and other services with single-file config mounts could keep running the previous file. Upgrades now detect and recreate services whose config changed. /health, /health/detailed, /ready and /livez now reach the API through the edge instead of returning the web UI.
  • OCSF installs wrote each audit event twice. On OCSF installs, audit events are now written once, to the OCSF table. The Audit page and search still show older events.

Known issues

These are fixed in v2.0.24. Upgrade to it.

  • Agent Lane tokens stop working about a minute after you create them. A background job that syncs tokens with ClickHouse doesn't receive the Agent Lane allow-list, and removes the token's access while Settings still shows it as active. Leave Agent Lane off until v2.0.24. Nothing else is affected.

  • scripts/nano upgrade fails its preflight on single-host installs. It stops with FAIL single runtime-config command failed (exit 78) and "preflight failed; no installation state was changed". This is a false failure in the upgrade check, not a problem with your install, and multi-host installs are not affected. Until v2.0.24, upgrade by re-running the installer from the extracted bundle:

    1. Take a backup: scripts/nano backup --key-file <operator.pem> --output <dir>.
    2. Extract the bundle over the install directory: tar -xzf nano-compose-enterprise-2.0.23.tar.gz --strip-components=1 -C ~/nano, then sign in to the registry.
    3. Run cd ~/nano && NANO_IMAGES_LOCK=images.enterprise.customer.lock ./install.sh. On the open edition, run ./install.sh.

    Data, rules and log sources are kept. Don't skip or edit the preflight check.

Upgrade notes

  • Required stop at v2.0.20. See above.
  • Upgrade with the release, not by editing image tags. On Docker Compose, use scripts/nano upgrade or re-run install.sh from the new bundle. On Kubernetes, use helm upgrade with the new chart. Both deliver the new Vector image and config refresh.

On this page