nano v2.0.12
nano v2.0.12 adds optional Self-run monitoring, fixes the Compose upgrade tool, checks managed PostgreSQL permissions first and fixes Helm and OTLP.
v2.0.12 includes everything in v2.0.11. It includes no security fixes.
To upgrade, follow Upgrade a self-hosted deployment. On Docker Compose, scripts/nano upgrade upgrades v2.0.10 and v2.0.11 to v2.0.12 directly.
Fixes and improvements
- Self-run monitoring (optional, off by default). Prometheus, Grafana with five dashboards, and alert rules for Docker Compose (
COMPOSE_PROFILES=monitoring), a role-split monitoring host, or Helm (monitoring.*), plus optional self-ingest of nano's health alerts into nano. See Monitor your Self-run install. - Log source health. A log source's health shows Fallbacks (24h) for events its parser fell back on, instead of counting them as successes, and Ignored claims for requests whose
X-Source-Typeheader doesn't match the token's source. See Routing model. - Docker Compose:
scripts/nano upgrade.- The dry run no longer fails with a false
registry-login ... not logged in to ghcr.io. Only nano's own application images need the registry login; the public images onghcr.io, Docker Hub and quay.io don't. Every image is still pulled by its signed digest before anything changes. - The dry run no longer fails with a false
ledger-clickhouseerror on installs with OCSF. The upgrader reads the OCSF migrations from the new bundle instead of a fixed list. - Upgrades from v2.0.10 and v2.0.11 to v2.0.12 work with
scripts/nano upgrade.
- The dry run no longer fails with a false
- Docker Compose: first database migration. On a fresh multi-server install, the migration could wait without end at "Applying open-init snapshot". It now refuses up front while PostgreSQL's synchronous standby isn't streaming, stops after 10 minutes (
NANO_PG_SNAPSHOT_TIMEOUT_SECS, 600) and reports what the session was waiting on. Nothing is half-applied. The installer boundsdocker compose upand prints the migration's last log lines if it fails. - Docker Compose: data-disk self-heal. The installer now installs the data-disk self-heal timer on a server whose Docker data is on a separate
nofaildisk. With passwordless sudo it does so itself, otherwise it prints the one command to run. SetNANO_DATA_DISK_HEAL=offto skip it. See Prepare the data disk. - Managed PostgreSQL: schema permissions. Before migrating, nano now checks that its user can create tables in schema
public. On Azure Database for PostgreSQL flexible server,publicbelongs toazure_pg_admin, so the first migration used to fail withpermission denied for schema public. The check names the user, the database, the schema owner and the command that fixes it. It blocks only when there are migrations to apply; an install with nothing pending logs a warning and keeps running. It applies to Azure, Amazon RDS and Cloud SQL. See Managed PostgreSQL requirements. - OTLP: retryable errors. When the check behind OTLP authentication can't reach the API, for example during a PostgreSQL failover, the ingest edge now answers
503withRetry-Afterinstead of500. OpenTelemetry exporters retry a503; they drop a500. With several app servers on Docker Compose, the check is retried on another app server. HEC and HTTP are not affected. - Helm.
- Vector's metrics port (9598) is reachable only from pods in nano's namespace. To scrape it from a monitoring namespace, set
vector.metrics.networkPolicyFrom. See Kubernetes: Network policy. - The ClickHouse operator image is pinned by digest, like nano's own images.
- Vector's metrics port (9598) is reachable only from pods in nano's namespace. To scrape it from a monitoring namespace, set
- Search: Rarity view. The Rarity view now finds domains and hashes in OCSF events.
- Health checks. After a ClickHouse pod restart, a stale distributed DDL entry now gives a warning instead of
UNHEALTHY: ch. - Other fixes.
scripts/nano support-bundlereports the install ID.- The message after the PostgreSQL servers' install no longer says to run
--record-pg-ha-identitywithsudo. - The HAProxy example enables the admin socket for the drain and rejoin hooks, and the bundle ships an example hook,
deploy/compose/examples/load-balancer/nano-lb-hook. Its advice to restrict the load balancer's SSH key to a forcedsudo socatcommand doesn't work; see Known issues. POST /api/rules/{id}/triggerreturns the rule ID and the time window it scanned.
- Documentation. Updated install, upgrade and high-availability guides for Docker Compose, and for Kubernetes on EKS, GKE and AKS.
Upgrade notes
- Helm: Vector metrics. If something outside nano's namespace scrapes Vector's metrics on port 9598, add its namespace with
vector.metrics.networkPolicyFrombefore upgrading, or the scrape stops on clusters that enforce NetworkPolicy.
Known issues
These are fixed in v2.0.13. Upgrade to v2.0.13 or later.
- Docker Compose, several hosts:
scripts/nano upgradekeeps each server's old configuration. It upgrades the images but keeps the nginx, HAProxy, ClickHouse, Keeper, PostgreSQL HA and monitoring configuration from the earlier install. So fixes in that configuration don't reach a cluster upgraded from v2.0.11. For example, the ingest edge still answers OTLP with500, not503, during a PostgreSQL failover. Workaround: after the upgrade, run each server's install command again, one server at a time, in the documented order, with its full role list. Each run renders the new configuration and recreates the services that use it. See High availability. A fresh v2.0.12 install is not affected. - Docker Compose, several app servers: after an app server reboots, the ingest servers can mark it
DOWN. Its internal API edge keeps the API container's old address. The UI,/readyzand SIEM Health look healthy; OTLP authentication and ingest checks run on the other app servers only. Workaround: after an app server reboots, runscripts/nano cluster-healthon each ingest server. Ifgw:app-api-proxyfails for that app server, restart itsapp-internal-edgecontainer. See High availability. - Docker Compose: the drain and rejoin hook advice. A forced SSH command that runs
sudo socatreplaces the hook's command and gets no input, so the drain silently does nothing. Use thenano-lb-haproxywrapper as the forced command instead. See Load-balancer drain and rejoin hooks. - Docker Compose: the post-upgrade SIEM Health check can compare the old report with itself, because no new report has run yet. Check SIEM Health yourself about 10 minutes after the upgrade.
- Docker Compose:
scripts/nano upgradeprints nothing until it ends when you run it over a plainsshwithout a terminal. Follow~/nano/upgrade-journal/*.jsonlon the first app server to see progress, or run it withssh -t.
nano v2.0.13
nano v2.0.13 makes detections, entities, prevalence and AI prompts correct on OCSF, makes OCSF the default for new installs and adds SCIM 2.0 provisioning.
nano v2.0.11
nano v2.0.11: a Helm security fix, Okta single sign-on, PostgreSQL 16 support, and install, upgrade and HA fixes for Docker Compose and Kubernetes.