nano v2.0.34
nano v2.0.34: SOAR case steps and case tags, script sandbox security updates, steadier detections, and fixes across Search, lookups, prevalence, cases, SOAR, rules and user management, with tighter access checks.
v2.0.34 includes everything in v2.0.33. It is a security and quality release. Its new feature is SOAR case steps, which let a flow work cases directly. It includes security updates for the script sandbox runtime and tightens several access checks, so we recommend that every self-hosted install upgrades.
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
None of these changes needs action from you.
- Script sandbox runtime. The release includes security updates for the script sandbox runtime that runs custom enrichments, collectors and agent lookups (Self-run). The sandbox also applies tighter limits on the memory and requests a script can use. Existing scripts keep working, and stored scripts are recompiled automatically.
- Webhook text is plain text. Untrusted text in Microsoft Teams card titles, summaries and facts renders as plain text, and invisible bidirectional and zero-width characters are removed from webhook text.
- Archiving a rule needs the delete permission. Deleting a rule archives it, so archiving now needs
detections:delete, the same as Delete. An archived rule opens read-only, with Unarchive as its only action. - Search cancel reaches only your own queries. The client
request_idsent with a search is now validated for format and length, and an id that could match another search's internal queries is refused with 400. - Demo visitors can't change AI settings. On demo deployments, AI settings, per-rule tuning settings and folder settings are read-only for visitors.
- Risk notables follow source restrictions. A risk notable is now scoped by the sources of the risk findings that built it. An analyst limited to some sources sees a notable when every contributing source is one they can read.
- No self-lockout. You can't disable or delete your own account, or remove your own Admin access, and nano refuses a change that would leave no active admin.
- Saved searches and dashboards you can't see answer 404. Editing, deleting or sharing a saved search, or deleting a dashboard, that you can't see now returns 404 instead of 403, so a private item's existence isn't revealed.
New
SOAR: case steps
A flow can now work cases natively, with no pack or connection. See Case steps.
- Create case opens a case, or reuses the run's case, the alert's case, or the case automation already opened for the same trigger event. Every flow fired by one alert works the same case.
- Add case note adds a Markdown note, and can include an earlier step's result, such as a search count.
- Set case severity, Set case status, Add case tag, Remove case tag and Assign case make the same changes a person makes.
- Each step runs as the flow's owner, through Cases, with the same permissions, lifecycle rules, timeline entries and audit as a manual change.
- Steps follow the flow's approval policy. A run triggered by AI, or a case step after a Judge step, always waits for a person, and the approver sees the exact change.
- Test runs never change a case.
Case tags
Cases have tags. Add and remove them in the case header, from a flow, or with GET/POST /api/cases/{id}/tags and DELETE /api/cases/{id}/tags/{tag}. A tag is up to 48 lowercase letters, digits or _ . : -, and a case holds up to 32.
Managed AI in demo mode
Demo deployments now use the managed AI provider, within the monthly AI allowance, so hunts and AI features work there.
Fixes
Search
sortandheadapply across the whole time range. A query such assort -count | head 10returns the top 10 for the whole search. Commands that look across rows, such asdedup,tail,reverse,join,append,eventstatsandtransaction, also run over the whole range.- Timecharts show every bucket. A timechart returns its complete set of buckets. Past 1,440 buckets nano widens the span and says so.
- Infinite scroll loads each page once, with no repeated rows.
- The Stats view shows every column, including columns added after
statsbyeval,rename,rex,lookuportable. When a result has more than the first page of rows, it says so and offers Load all (up to 10,000). - Export works for timecharts and top/rare as CSV.
- Cancel stops every query a search started.
If you have saved searches or dashboards that use sort with head, timechart, lookup … OUTPUT or the prevalence features below, or that end in table, fields or rename with no sort, consider re-running them once. The same goes for results you gathered by scrolling.
Lookups and prevalence
lookup … OUTPUTfills the listed columns, including withASon the key.lookup … OUTPUTreturns only the columns you list. It no longer adds the key column unless you name it inOUTPUT. WithoutOUTPUT,lookupstill adds every column.- Lookup tables keyed on any column read back correctly, including keys named
user,grouporkey. - Editing a lookup table cell updates the row in place.
| prevalence enrich=true(Check prevalence in Search) shows host counts for artifacts first seen in the last few minutes.- Search events on the Prevalence page finds events for process hashes, DNS and URL domains, and source IPs, over the page's time window.
Cases
- Choosing Closed in the case header opens the close flow, which asks for the reason and disposition.
- Every new case gets a notebook, including unassigned cases, so you can add notes right away.
- Analysts see SLA targets on the case page without the settings permission.
- Analyst notes render Markdown.
- Grouped cases on OCSF data are titled with the grouping host, user or IP.
SOAR
- The SOAR catalog verifies and updates again. Packs and updates from the content channel apply as before.
- Approved decision gates in playbooks run their action. A playbook action gated on a decision (
when: d1 = yes) now runs when the approver picks a matching option. It still waits for its own approval. - Flows triggered by AI events wait for a person before every action, notification and case step. See Behavior changes.
- Testing an approval followed by a condition takes the condition's default branch and shows not evaluated, instead of failing.
- A flow with an action that has no connection can't be saved as live, and a trigger refused for that reason is shown as Skipped with the reason.
- Run test with a real alert accepts the alert's
alert_ID. - The test run trace shows each step's recorded state and sensible timings.
Detections and rules
- A rule that keeps crashing the jobs service backs off. It shows as failing in rule health, and retries after a growing delay instead of restarting the service again and again. Rules that were running alongside it when it crashed aren't counted as failing.
- Very large event rules are bounded per run. A rule whose results are plain matching events evaluates at most 2,000 events per run by default: the newest 2,000 plus the newest event for every distinct entity, so every entity still alerts. The run is marked PARTIAL. See Bounds per run.
- A deleted rule stays deleted. It no longer appears by ID, can't be triggered or promoted, and isn't offered again by a repository sync. You can still import it again on purpose.
- Linked repository rules show DRIFT when edited locally.
- MITRE coverage shows rules in review and disabled rules, and refreshes on every visit.
- Risk notables open for analysts, and the alert page renders them.
Users, notebooks and settings
- Deleting a user or group that a case handoff or escalation was addressed to works, and keeps the handoff or escalation. A pending handoff that loses its last recipient is canceled, the sender is notified, and its SLA timer resumes.
- Agent model changes in Settings reach every nano service within about 30 seconds, with no need to save a provider again.
- Deleting a user keeps case evidence. Their case notes and handoffs stay, shown as by "deleted user", and their notebooks that others use are handed to a teammate. Only private notebooks nobody else used are removed.
- Saved queries can be edited and deleted from the palette, and keep their description.
- A new notebook shows in the list right away, and Create New Notebook in the palette opens the dialog.
- With an active notebook, every completed search is captured.
- Pages that read-only users and analysts can open no longer make calls that return 403. ReadOnly users can browse rule repositories.
- A fresh install no longer logs a 404 on every page for the SIEM health report.
- pivt's page card names Lookup tables, IOC lists and Retro hunt correctly.
Install
- Ctrl-C at an interactive
install.shprompt exits cleanly and leaves the host unchanged on newer bash releases (5.3 on Linux), where it could wait at the prompt. Typed answers are never lost.
Log sources
- Deleting several log sources reloads Vector once. Deletes made close together share one reload, and
POST /api/log-sources/bulk-deletedeletes up to 500 sources in one call. Deletes have a longer timeout, so a slow reload no longer returns 504. - The
log_source_deletedaudit entry has the source name and itslsrc_ID.
Behavior changes
- Archiving a rule needs
detections:delete. - SOAR flows triggered by AI events wait for a person. A run triggered by a hunt-finding alert, the AI triage verdict, a case the AI opened (or one seeded from a hunt finding), or a case status change the AI made is treated as AI-requested. Every pack action (including
readactions and steps set to automatic approval), notification, Set hunt finding disposition and case step in it waits for approval. Live flows on these triggers that ran unattended before the upgrade now pause on Automation → Approvals. - Lookup row IDs are strings.
_row_idinGET /api/lookup-tables/{name}/rowsandrow_idsin thePOST …/rowsresponse are now decimal strings, so large IDs keep their exact value. Requests still accept a string or a number. Updating or deleting a row ID that doesn't exist answers 404. - Detection runs are bounded. Plain event rules that match more than
DETECTION_MAX_MATCHES_PER_RUNevents (default 2,000) in one run evaluate the newest 2,000 plus the newest event of every distinct entity, and the run is marked PARTIAL. If there are too many distinct entities to cover, a DETECTION COVERAGE INCOMPLETE alert fires at the rule's severity. Aggregation and risk rules are unchanged. Stored match rows are capped at 512 KiB, and match counts stay exact. - Archived rules return 404 by ID.
GET /api/rules/{id}andPOST /api/rules/{id}/triggeranswer 404 for a deleted or archived rule. Passinclude_archived=trueto read it. lookup … OUTPUTno longer outputs the key column unlessOUTPUTnames it. Review saved searches, rules and dashboards that read the key column afterlookup … OUTPUT.- Approved decision gates in playbooks run their action. Playbook automation queued before the upgrade keeps the old behavior; start a new playbook run on the case to pick up the fix. Actions behind a chained decision, or behind a
when:value the decision doesn't offer, stay manual and show the reason. - SOAR approval conditions must read
.choiceor.approved. Validation refuses a condition that compares the whole approval result. Such a condition never matched. - Run test with a real alert takes
alert_IDs, and still accepts a bare UUID. - Client
request_idvalues are validated. A search whoserequest_idhas an unexpected format, length or suffix is refused with 400. nano's own clients are not affected. - Deleting a user keeps case evidence. Notebooks are reassigned and entries show "deleted user". If no one can take over a shared notebook, the delete is refused with 409.
- Air-gapped installs refuse Git repository sync. Creating or syncing a rule, parser or playbook repository answers 409 on an air-gapped install. Use the offline bundle import.
- Users can't disable or delete their own account or remove their own Admin access (403
cannot_modify_self), and a change that would leave no active admin is refused (409last_admin_required). GET /api/siem-health/reports/latest?allow_empty=truereturns 200 with JSONnullwhen no report exists. Without the parameter it still returns 404.
API changes
GET /api/rules/{id}/matchesaddsevents_truncatedfor a match whose stored events were cut down to fit.CaseHandoffandCaseEscalationaddtarget_removed_at, set when the recipient user or group was deleted.- Lookup row IDs,
include_archived,request_idvalidation and the 404 answers are described under Behavior changes.
Upgrade notes
- Required stop at v2.0.20. See above.
- Check flows triggered by AI events. Live SOAR flows triggered by hunt-finding alerts, the AI triage verdict, AI-opened cases or AI case status changes now wait for approval at every action, notification and case step. Make sure someone with
soar:approvewatches Automation → Approvals. - Migrations run automatically. New PostgreSQL migrations apply during the upgrade: core 392, 394, 395, 396 and 397, and on Self-run 9000136 to 9000140. They add saved search descriptions, the rule repository read permission for ReadOnly, rule drift baselines, detection crash tracking, case tags, one notebook per case, and the user and handoff delete changes. No long table locks are expected. Reference changes are added without a table scan and checked afterwards without blocking reads or writes. There are no ClickHouse migrations.
- Duplicate case notebooks are merged. If a case has more than one notebook, the upgrade moves the newer notebooks' entries into the oldest one and archives the rest. Entry order is unchanged.
- Risk notables created before the upgrade keep their previous scoping, so only unrestricted users see them. New notables are scoped by their sources from the next rule run.
- Script sandbox recompiles. Stored custom scripts are recompiled on first use after the upgrade. The first run of each may take a little longer.
- Single-server Compose keeps Vector's working files outside its watched folder, in
/app/data/vector-work. The upgrade applies this for you. - Empty grouped case titles created before the upgrade keep their title. New cases get the full title.
Upgrading from v2.0.32
v2.0.33 changed the hunt escalation outbox. On a rolling upgrade from v2.0.32 (several hosts or Kubernetes), a server still running v2.0.32 can fail to queue hunt escalation notifications until it restarts on the new version. Finish the rollout without long pauses between servers. Single-server installs restart every service together and are not affected.
Related docs
Upgrade a self-hosted deployment
Check which nano version you run, then upgrade an Open or Self-run deployment on one host, several hosts or Kubernetes to the latest release.
nano v2.0.33
nano v2.0.33: an nginx security update, run caps and a circuit breaker for SOAR, team and private escalation queues and a duplicate gate for Hunts, one risk score per entity, and search guards based on query cost.