nano v2.0.8
nano v2.0.8 includes security fixes, an enforced ingestion pause, a more reliable CrowdStrike FDR source, and install and high-availability fixes.
v2.0.8 includes everything in v2.0.7.
To upgrade, follow Upgrade a self-hosted deployment.
Security
This release fixes issues found in an internal security review. We don't publish exploit details. To report a vulnerability, see Support.
- First-admin setup needs the setup token. Creating the first administrator now requires the setup token the installer generates. Without one, only a direct call on the server itself is accepted. See Upgrade notes.
- SSO trust changes need administrator authority. Creating or enabling an SSO provider, or changing its issuer, client ID or group claim, requires full administrator authority. So does widening who can sign in: turning on sign-up for mapped groups, adding an Entra tenant, trusting a new verified domain or adding a Google service-account key. Changing the issuer signs out and unlinks the accounts linked through the old one. See Single sign-on and Upgrade notes.
- Credentials follow their destination. Changing where a collector, webhook or AI provider sends data asks for its credential again.
- Enrichment lane. The push-enrichment ingest lane accepts only enrichment credentials.
- Parser tests run in a separate, resource-limited worker process.
- Search links from outside nano load their query and ask before running it.
- Access-scope, integrity and release-provenance hardening.
Highlights
- The ingestion pause is enforced. When ingestion is paused, manually or at the emergency disk-pressure level, nano now refuses new events at the ingest edge and holds back writes from Vector and Tenzir until it resumes. Senders keep their data and retry. Settings > Storage shows whether ingestion is paused, why and since when, and lets an admin pause or resume it. See Disk pressure.
- More reliable CrowdStrike FDR. The source downloads each file completely before reading it and retries a dropped download. If a batch is delivered again, files it already delivered are skipped, so a dropped download no longer duplicates events. See CrowdStrike FDR.
- Install fixes for multi-node deployments. Health and readiness checks use each server's configured bind address, and the deployment planner, preflight checks and
nano-deployskill have fixes from a full-HA dry run. See Deploy with your AI agent. - High availability. Revoked sessions stay revoked across a Valkey failover. See Test failover.
Upgrade notes
- Setup token. Upgrading adds
NANO_SETUP_TOKENto.env(on Kubernetes, thesetup_tokenkey of the generated secret). You need it only to finish an interrupted install on the setup page; an install that already has an administrator is unaffected. - Rules that set
source_type. For a detection rule that writessource_type, matched events show that value asrule_source_type, and alerts keep the source of the events the rule searched. Such a rule's event hash changes once. - Editing a deployed source now needs the same permission as deploying it. This includes adding, changing or reordering its routing rules. Changing only its description still needs only edit permission.
- SSO and older admin keys. SSO provider create, enable, and issuer, client ID or group claim changes now need every current administrator permission. So does widening admission (sign-up for mapped groups, a new Entra tenant, a new verified domain or a Google service-account key). Older admin API keys, and custom roles copied from Admin before newer permissions existed, may get 403 there. Recreate the key, or add the current Admin permissions to the role. Before you change an issuer, make sure a local administrator account can sign in.
- Logout on high-availability installs. Logout returns 503 only when PostgreSQL can't record the revocation. If Valkey can't confirm it, logout still succeeds, and every server enforces the revocation from PostgreSQL within about 5 seconds.
- Alerts from rules with a subsearch. If you restrict any sources, some alerts are now hidden from users who can't read every restricted source. This applies to a rule that uses a subsearch (
append,joinorIN [...]) when its results carry no per-event source, for example afterstats. Users withsource_scopes:view_allstill see them. See Source visibility. - Push-enrichment lane.
/enrichmentnow accepts only the platform credential (NANO_PLATFORM_INGEST_TOKEN) or an install token that is not bound to a source. Records sent with a per-source ingest token are dropped as ingestion errors, and the sender still gets 200. Compose and Helm installs already give nano-api and nano-jobs the platform credential. If yours run without it, and your install token is bound to a source, identity sync records are dropped until you set it. - CrowdStrike FDR on Helm or Kubernetes. Every Vector pod runs the FDR source. Keep Vector at one replica (
vector.replicas: 1andvector.hpa.enabled: false, the defaults) while you collect FDR. With more pods, a re-delivered batch can reach a pod that never saw it, and its events are stored twice. The record of delivered files lives on the pod's ephemeral volume, so replacing the pod (an upgrade or a node drain) can also duplicate a batch that was cut off. On Compose, the record is on a persistent volume. - Migrations are forward-only and run before the new services start.
- Upgrade with
install.sh, as described in Upgrade a self-hosted deployment.