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.
v2.0.26 includes everything in v2.0.25. It is a security release. Every self-hosted install should upgrade.
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.
Security
- Unusual event data could stop detection rules from alerting. An event value containing certain characters, such as U+0000, could make every rule that outputs that field fail. A rule without a lookback then stayed stuck, and a rule with a lookback lost the matches in its window. Health still showed ok. Event values are now made safe before they are stored with an alert, one bad event no longer stops the rest of a batch, and a rule that keeps failing records a coverage gap and shows as failing in rule health, SIEM Health and
/health/detailed. Details: GHSA-x5wv-wr6v-4fvx (Medium, CVSS 6.5). Thanks to @bhaswanthc for the report. - Searches written by AI agents can't reach the network or write data. pivt, investigation chat, shadow investigation and nano Desktop agent searches run with a read-only command set.
inputlookup url=,aiandoutputare refused for them. Admins can allow specific commands under Settings, Search. - The lookup URL allowlist is enforced when set. With
INPUTLOOKUP_ALLOWED_DOMAINSset,inputlookup url=only fetches listed hosts and their subdomains, checked before any DNS lookup and on every redirect. With it unset, behavior is unchanged.
Fixes
- One bad event no longer holds up real-time detection. A real-time rule's failure is now isolated to that event. Data errors are recorded as coverage gaps with a reference alert; a ClickHouse or PostgreSQL outage holds detection until it recovers and raises an alarm, instead of dropping events.
- Searches handle invalid text safely. Functions such as
substr,reverseandbase64_decodeno longer produce text that breaks a search, and conversions such asto_timestamp,to_dateandtonumberreturn empty for bad values instead of failing the rule. - Detection catches up after an outage. A ClickHouse restart or outage no longer makes rules skip time. Rules wait and catch up.
- Evidence Pack runs at default sizing on smaller enterprise hosts. The memory a pack needs now depends on its period, and enterprise sizing gives the API at least 1 GiB.
- Restore works on Debian and Ubuntu. The free-space check no longer fails under the default
awkthere. - Kubernetes upgrades no longer stall when old and new API pods briefly serve different Vector configurations.
- Kubernetes monitoring: a Prometheus running in the
monitoringnamespace can scrape Vector and ClickHouse metrics without extra settings.
Upgrade notes
- Required stop at v2.0.20. See above.
- New rule health fields. The rules API now includes
consecutive_failures,failing_since,last_errorandlast_error_at. Seedocs/administration/detection-execution-health.mdin the release bundle.