nano v2.0.27
nano v2.0.27: disk protection on by default, ClickHouse storage tiering, egress controls, and updated libraries and base images.
v2.0.27 includes everything in v2.0.26. It is not a security release.
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.
Highlights: disk protection and storage tiering
- Ingest pauses before the disk fills. At 92% disk use, ingestion pauses by default and the ingest edge answers
503withRetry-After. Vector's buffer can no longer fill the disk. Ingestion resumes by itself below 80%. - Disk eviction keeps a retention floor. The disk-pressure service never drops days newer than the floor. When only those are left, it alerts and pauses ingestion at 90% instead. It drops data only when ClickHouse caused the growth.
- The audit trail survives eviction. Audit rows are copied to a new
audit_archivetable, and checked, before their day is dropped. The Audit page still shows them. - Disk health with a forecast. SIEM Health and
/health/detailedshow each disk's growth rate and hours until full. At 90%, or 24 hours from full, admins see a banner. At 95%, a red banner stays up and alerts repeat every hour. - Retention you can size and set. With a planned volume,
install.shand Helm say how many days the disk holds and refuse a new install under 7 days. Set retention withNANO_RETENTION_DAYS,nano.retention.daysor Settings, Storage. - ClickHouse storage tiering (Self-run). Move older data to S3, an S3-compatible store, GCS or Azure Blob as the local disk fills. Use
scripts/nano tieringon Docker Compose orclickhouse.coldStorageon Kubernetes. While tiering is on, disk pressure moves old days to the cold tier and never drops them.
Egress and AI
- Proxy-required mode. With
NANO_EGRESS_PROXY_REQUIRED=true, outbound clients never connect to a public address directly.DESTINATION_EGRESS_PROXY_URLgives webhooks, AI providers andinputlookup url=a proxy. Both are off by default. - Kubernetes egress NetworkPolicy.
networkPolicy.egress.enabled: trueadds a default-deny egress policy. It needs a CNI that enforces egress. See the known issue under Upgrade notes, fixed in v2.0.28, before you turn it on. - Lookup URL allowlist in Settings. Admins manage the hosts
inputlookup url=may fetch under Settings, Lookup URLs. Each fetch is audited. - pivt case actions (Self-run). If pivt read log data or notebook entries, it can't close, dismiss, downgrade or reassign a case until an analyst approves. Each action is attributed to pivt and can be undone.
- AI-search refusals in the audit log now name the user and client IP.
- Parser saves no longer fail as invalid VRL when the validator is busy. They wait, then return a retryable
503. - Agent Lane (Self-run): with the listener off, Settings shows it off and its tokens as inactive.
Libraries and platform
- Toolchain and images: Rust 1.100, Node 22 LTS for the web build, and Debian 13 (trixie) runtime images. Building from source now needs Rust 1.100.
- Rust libraries: sqlx 0.9, clickhouse-rs 0.15, aes-gcm 0.11, hkdf 0.13, ed25519-dalek 3, jsonwebtoken 12, totp-rs 6, utoipa 6 with Swagger UI 10, zip 9, tower-http 0.7 and sqlparser 0.64. Stored credentials, licenses and authenticator apps work unchanged.
- Components: Tenzir 6.22, etcd 3.6.16, redis_exporter 1.97, and updated web and desktop UI libraries. Images are pinned by digest.
Fixes
- Rule tuning stats are much faster: about 34 seconds to 0.2 seconds on 16.8 million findings.
- The ingest watchdog restores its state after a restart.
- Alert cooldowns work for long entity values, such as long command lines.
- Deeply nested events are kept as JSON text past 24 levels. SIEM Health reports events Vector drops.
- Background searches report the real cause when their summary can't be saved.
- Old in-app notifications are pruned: read ones after 30 days, all after 90.
- Kubernetes node alerts fire with kube-prometheus-stack, and new rules watch PVCs.
Upgrade notes
- Required stop at v2.0.20. See above.
- Auto-pause is on. To change the level, set
DISK_PRESSURE_PAUSE_THRESHOLD(Compose.env) ornano.diskPressure.pauseThreshold(Helm), between the high watermark and0.99. To turn it off, setDISK_PRESSURE_PAUSE_INGESTION=falseornano.diskPressure.pauseIngestion: false. - Retention floor. By default the floor is 7 days, or one day less than your oldest data if that is smaller. To set it, use
DISK_PRESSURE_MIN_RETENTION_DAYSornano.diskPressure.minRetentionDays. - Custom Tenzir pipelines. The shipped
tenzir/pipeline.tqlruns unchanged on Tenzir 6.22. An edited one can break: legacy operators are gone, HTTPS sinks check certificate hostnames, and listeners check client certificates only againsttls.client_ca. When you upgrade withinstall.sh, an edited pipeline the new package would replace is saved totenzir/pipeline.tql.bakand the install stops once.scripts/nano upgradefrom v2.0.26 does not do this check, so copy an editedpipeline.tqlsomewhere safe before you run it. - Network pin and ClickHouse lockout check.
scripts/nano upgradeon one host now pins nano's network, which restarts the stack once. If an outside ClickHouse writer would be locked out, it stops and names theNANO_CLICKHOUSE_INGEST_ALLOWED_NETWORKSvalue to set. To continue anyway, pass--ack-clickhouse-network-lockout. - Monitoring RAM check.
scripts/nano upgradenow also updates the monitoring stack, and its preflight counts 1 GiB for it. If that fails, add RAM or removemonitoringfromCOMPOSE_PROFILES. - These two
scripts/nano upgradechanges apply to upgrades from v2.0.27 onward, because the upgrade runs the installed release's upgrader. - Retention repair. Earlier releases could cut the delete TTL of
logs,signalsandingestion_errorsto 30 days when cold storage settings were set. The upgrade restores it once and raises a SIEM Health notice. Data already removed can't be recovered. - Migrations run automatically. New PostgreSQL and ClickHouse migrations apply during the upgrade.
- Helm disk check. With
sizing.plannedGBPerDayset (every sizing preset sets it), the chart refuses a first install whose ClickHouse disk holds under 7 days. Ahelm upgradeof a running release only warns. A smaller disk than the preset assumes, such as a 150Gi evaluation disk withvalues-standard.yaml, still upgrades.helm templatewithout--is-upgrade,helm install --dry-runand Argo CD render as a first install, so they refuse it. Setsizing.allowUndersizedDisk: truein the values they use, raiseclickhouse.storage.size, or turn on the cold tier. See Upgrades. - Known issue, fixed in v2.0.28: an interactive
install.shstops at the storage tiering question. Answering "Enable cold storage tiering?" in a terminal fails withio.UnsupportedOperation: File or stream is not seekable. Fixed in v2.0.28. On v2.0.27, runNANO_TIERING=off ./install.sh, then turn tiering on later withscripts/nano tiering configureonce you are on v2.0.28. Unattended installs (no terminal) are not affected. - Known issue, fixed in v2.0.28: the egress NetworkPolicy breaks DNS on clusters with a node-local DNS cache. That includes stock GKE (NodeLocal DNSCache) and AKS with LocalDNS. Its DNS rule admits only kube-dns pods. The policy applies to the running pods too, so turning it on during an upgrade stops every nano pod. Fixed in v2.0.28, which admits the DNS ports to any destination by default. On v2.0.27, leave it off. Clusters without a node-local cache on the host network (not GKE Dataplane V2 with Cloud DNS or Autopilot) can instead set
networkPolicy.egress.dns.toas described in Egress NetworkPolicy.
nano v2.0.28
nano v2.0.28: the interactive install.sh fix, one ClickHouse user per workload, ingest no longer refused at the hourly query quota, and upgrade and Helm fixes.
nano v2.0.26
nano v2.0.26: a security and reliability release. Detection rules keep alerting on unusual event data, failures are visible in health, and AI-written searches can't reach the network.