nano v2.0.30
nano v2.0.30: Vector native port 6000 always requires mutual TLS, with a certificate per sender. The installer checks the host before it asks anything, shows a review screen, and Self-run installs and upgrades log in to the registry themselves.
v2.0.30 includes everything in v2.0.29. It is a security release: the Vector native listener on port 6000 now requires a client certificate from every sender.
To upgrade, follow Upgrade a self-hosted deployment and the Upgrade notes below. v2.0.20 is a required stop: installs on a release older than v2.0.20 upgrade to v2.0.20 first.
Security
- Vector native port
6000requires mutual TLS. The Vector protocol has no token, so a client certificate is now the credential. Every connection must present one issued for your install, or the TLS handshake is refused before any event is read. Up to v2.0.29, the port was plaintext with no credential check. It was bound to127.0.0.1by default on a single server, and on a role-split cluster no ingest server published it. Now a single server listens on every interface, like8080and8088, andNANO_VECTOR_NATIVE_BIND_ADDRESSnarrows it to one address. On a role-split cluster each ingest server publishes it on its private cluster address; setNANO_VECTOR_NATIVE_BIND_ADDRESS(for example0.0.0.0) for senders outside the private network.- One certificate per sender.
scripts/nano vector-client-cert <name>issues it and prints a ready-to-paste Vector sink.--csrkeeps the key on the sender.--listshows each sender asactive,expiring,expiredorrevoked, and--renewre-signs a sender with no restart on nano. - Revoking one sender never affects the others. Each sender has its own trust anchor.
--revoke <name>removes only that anchor and restarts Vector. No other sender needs new files. - Fails closed. If a TLS file is missing or unreadable, the port refuses every sender instead of falling back to plaintext. The other ingest lanes stay up.
- Backups.
scripts/nano backupnow includes.nano-pki, and restore puts it back. - Kubernetes. The chart creates the listener CA and certificate, and keeps a sender anchor bundle that starts locked. See Vector-native client certificates.
- One certificate per sender.
Installer
- Host minimums are checked first. Before any question or
.envwrite, the installer refuses a single server with fewer than 2 vCPU, less than 8 GB RAM or less than 40 GB free on Docker's data disk. A host where ClickHouse shares the server with the other services and the enrichment feeds are on needs 16 GB. A host that already runs nano only gets a warning.NANO_SKIP_HOST_MINIMUMS=1turns the refusal into a warning. Smaller floors apply to role-split hosts without ClickHouse. nano-api is never sized below what it needs to start. - A review screen before anything is written. Every question is asked first, then a numbered review screen lists the answers. Enter installs, a number changes that answer, and
qquits.back(or<) at any prompt goes to the previous question. Nothing is written before you confirm: no.env, no secrets, no registry login, no container. Unattended runs skip the review. - Re-runs keep what you set. A
NANO_COMPOSE_OVERLAYSyou pass (for example the TLS overlay) is recorded in.env, so re-runs andscripts/nano upgradekeep HTTPS.NANO_BASE_URLon a re-run changes the base URL. A re-run that changes release says it upgrades and that database migrations can't be undone. In a terminal it asks first.nstops with exit code3, and Enter or end of input with exit code4; nothing is changed.NANO_CONFIRM_UPGRADE=1answers yes in advance. Runs with no terminal proceed as before. With no.envbut the old data volumes present, the installer stops and explains how to restore the old.envinstead of starting over with new passwords. - Self-run installs log in to the registry themselves. Enterprise installs now use
images.enterprise.customer.lockby default, and the installer logs in with the license values in.envbefore it pulls.scripts/nano upgradedoes the same. The login lives in a private, temporary Docker configuration that is removed afterwards: your own~/.dockeris never written, and an existing login of yours for the registry is left alone. You no longer setNANO_IMAGES_LOCKor rundocker loginby hand. Air-gapped installs with images already loaded make no registry call. - Clear errors before the long steps. The installer stops early, with the fix, when:
- the TLS overlay is on but the certificate is missing, expired, has a passphrase or doesn't match its key. On an existing install, where nginx may already be serving it, an expired or unreadable (for example root-owned) certificate only gets a warning.
NANO_SKIP_TLS_PREFLIGHT=1skips the check; python3is missing (it names the install command for your distribution);- your user isn't in the
dockergroup, or.envbelongs to root; - port
80,8080or8088is already in use (it names the process or container). Re-runs no longer report nano's own ports as in use.
- the TLS overlay is on but the certificate is missing, expired, has a passphrase or doesn't match its key. On an existing install, where nginx may already be serving it, an expired or unreadable (for example root-owned) certificate only gets a warning.
- Better failure reports. When a service fails to start, the installer shows ClickHouse's full error, logs from every failing or restarting service, and how many times nano-api was killed for running out of memory. A
NANO_API_MEM_LIMITset below the minimum is named with where it was set. - Upgrade fixes.
scripts/nano upgradeno longer stops with exit code 141 on installs with many images, and--resumeno longer says the install has no online license.
Fixes
- HTTPS installs are healthy. With the TLS overlay, nginx's health check probed plain HTTP, which only redirects, so nginx stayed
unhealthyandscripts/nano upgradecouldn't pass its edge check. The check now probes HTTPS on443. - EKS egress policy. With
networkPolicy.egress.enabled: trueon EKS, api, search, jobs and the connector runner stayed inInit. The VPC CNI agent matches thekubernetesService ClusterIP, not the API addresses. The chart now adds it withnetworkPolicy.egress.kubeApi.serviceIP(autoby default). See Egress NetworkPolicy. - Search autocomplete keeps hyphenated values. Choosing
web-app-jsonformetadata.log_source=web-app-jsonno longer rewrites it toweb-app-"web-app-json". - The rule test panel tests the rule you are editing. It used to keep the query from when it opened. It now follows the editor. A query you change in the panel is labeled, and Apply to editor asks before it replaces your rule.
Upgrade notes
- Required stop at v2.0.20. See above.
- No registry login step from v2.0.30. With an online license, the v2.0.30 installer and
scripts/nano upgradelog in to the registry themselves. You can skip the login step in Upgrade a self-hosted deployment. Offline licenses still log in by hand or use the image bundle. NANO_IMAGES_LOCKis no longer needed on Self-run. An install that recorded the internal image names after a failed v2.0.29 install, with no container created for them, switches to the customer registry (same digests) and says so.- Host minimums. An existing install below the new minimums gets a warning, not a refusal.
- A low
NANO_API_MEM_LIMITis refused. If you pinnedNANO_API_MEM_LIMITin.envor the environment below 640 MiB (Self-run) or 384 MiB (Open), the installer now refuses instead of starting nano-api too small to boot. Raise it or remove the line so the installer sizes it. - Re-runs ask before an upgrade. A
./install.shre-run in a terminal that changes release now asks[y/N]. Declining exits3(n) or4(Enter or end of input). If you drive the installer through a terminal from a script, setNANO_CONFIRM_UPGRADE=1. - A process holding port 6000. Before the rollout,
scripts/nano upgradechecks that no other process on the host holds port6000on the address nano binds, and refuses with the cause if one does. Stop that process or setNANO_VECTOR_NATIVE_BIND_ADDRESS, then run the upgrade again. - Migrations run automatically. New PostgreSQL and ClickHouse migrations apply during the upgrade.
Vector native senders
If nothing sends to port 6000, there is nothing to do. To check before you upgrade, search source_type=vector_native over the last 24 hours. On a single server, an install.sh re-run also reports that count.
Otherwise, after the upgrade:
-
On the nano server (role split: the first app server), issue one certificate per sender:
cd ~/nano scripts/nano vector-client-cert edge-dc1This writes
edge-dc1.crt,edge-dc1.keyandca.crtto./nano-vector-edge-dc1/and prints the sink config. Use--csr <file>to keep the key on the sender. -
Role split only: copy the first app server's
.envto every server and update every ingest server, so they trust the new sender. See Vector native (port 6000). -
Copy the three files to the sender and add the TLS settings to its Vector sink:
[sinks.nano] type = "vector" inputs = ["<your source>"] address = "nano.example.com:6000" version = "2" tls.enabled = true tls.crt_file = "/etc/vector/nano/edge-dc1.crt" tls.key_file = "/etc/vector/nano/edge-dc1.key" tls.ca_file = "/etc/vector/nano/ca.crt" -
Restart the sender's Vector and check that events arrive.
- Port
6000is now reachable from the network. On a single server it moves from127.0.0.1to every interface; to keep it on one address, setNANO_VECTOR_NATIVE_BIND_ADDRESSin.env(for example127.0.0.1if every sender runs on the server). Role-split ingest servers publish it for the first time, on the cluster address. Docker publishes it, so UFW doesn't filter it. Allow it only from your senders in your cloud firewall or Docker'sDOCKER-USERchain. See Firewall. - Senders through NAT or a load balancer need that name or address in the listener certificate: add it to
NANO_VECTOR_NATIVE_SANSin.envand re-runinstall.sh. - Plaintext lab clusters (
allow_plaintext_private_network) have no listener: the port refuses every sender. - Kubernetes with
vector.native.enabled: true: senders without a client certificate are refused from the upgrade on. Follow Vector-native client certificates.
Full details are in Vector native (mutual TLS).
nano v2.0.31
nano v2.0.31: a hotfix. Fresh role-split installs no longer return 503 when a slow ClickHouse health check fails, scripts/nano upgrade no longer fails its SIEM Health check on errors the rollout itself caused, and the installer's re-run upgrade prompt defaults to yes.
nano v2.0.29
nano v2.0.29: a security release. A package extract no longer widens ClickHouse's network allow-list or resets its sizing, busy single ingest hosts upgrade without pausing senders, and image bundles work on registry installs.