Application changelog

This page records operator-visible controller behavior released through the DonkeyFleet application image. Helm chart versions are tracked separately in the Helm chart changelog.

1.3.0

• Redesigned volume exclusions. Removed the regex ignore-rule mechanism; foreign-owned, scratch, and temporary volumes are now hand-picked into named exclusions — each with a name and a reason — on Policy → Exclusions. Search discovery, stage a bulk list, and add; add more volumes to an existing name to grow it. Excluded volumes are listed grouped by exclusion name with each volume's provisioned and physical size, and Include returns one to candidacy. Clones stay excluded structurally and appear as a read-only group.
• Excluded volumes now auto-retire when their source volume is deleted: once a complete reconcile no longer sees the volume, DonkeyFleet removes the exclusion entry and its stale candidate, keeps the named exclusion for reuse, and raises an informational notification. A partial observation or an unreachable cluster never triggers it.
• Added an Excluded capacity bucket. The Capacity page's Overview shows Protected, Unprotected, Orphaned, and Excluded exposure side by side — volumes, provisioned, and physical — and the home dashboard surfaces an excluded-volumes total.
• Moved clones out of the Approvals backlog; they are surfaced under Exclusions instead.
• Renamed the runtime-safety controls for clarity: the kill switch is now Emergency stop / Release emergency stop, and live apply is Go live / Return to dry-run / Clear manual override. The endpoints and the persisted force-dry-run setting are unchanged.
• Rewrote the runtime-safety pane in plain language, with a per-button explanation beside each control.
• Removed the duplicate "Unread notifications" card from the home dashboard; "Awaiting approval" already covers it.
• Reconcile logging now reports the excluded-candidate count alongside observed candidates and relationships.
• Overhauled the Notifications page: advisories are grouped into collapsible per-volume panes with a per-severity breakdown, a single date-range filter, pagination, and a Source/Destination subject label on each card.
• Fixed the inventory rows-per-page control leaking across tabs; each tab keeps its own page size and the Exclusions tab no longer shows the control.
• Fixed the home "Protected volumes" card to open the Capacity page's by-automation-profile view.
• Fixed the Approvals table search so the term also filters with "All automation profiles" selected (previously it only worked after choosing a specific profile); applies to the Pending, Backlog, Apply, and Cleanup tables.
• Simplified volume-group match operators to Contains, Starts with, Equals, and Not equals — dropping Regex and Ends with (error-prone, and a poor fit for Trident's UID-suffixed names). Existing rules keep evaluating.
• Fixed the Notifications page count: the header now counts unread notifications, matching the nav badge, instead of counting volume groups — the two disagreed whenever one volume carried more than one unread advisory.

1.2.0

• Show the running DonkeyFleet release in the console footer on every page.
• Bounded the autonomous grow step to a 5% or 10% choice (default 5%), enforced; the full resize to target utilisation still always requires human approval.
• Auto-resolve a destination-fill ("near full") notification, and the now-moot resize-approval advisory, when the fill episode closes — so neither lingers after the fill clears.
• Escalate autonomous-grow "riding": warn after repeated autonomous grows without an approved resize, and raise an urgent alert when the cumulative-growth ceiling is reached (approve the resize, or the volume keeps filling).
• Fixed the relationship notification deep link to open the relationship detail page, and removed the redundant button from the near-full advisory.
• Grouped the automation-profile form into General, Orphan cleanup, Fill & grow, and Transfer sections.
• Added protected totals to the Capacity page (volumes, provisioned, and physical across all automation profiles) and a total-protected-volumes card to the home dashboard.
• The "near full" alert now describes the bounded autonomous grow in Auto mode, instead of only pointing at the approval-gated resize.

1.1.0

• Detect a replaced source volume (same name, new UUID) as source_replaced drift instead of an orphan: the relationship stays drifted, is excluded from remediation and cleanup, is never auto-remediated, and raises a warning. A repair path reclassifies any relationship an earlier build had marked orphaned, cancelling a stale approved cleanup.
• Added a per-automation-profile capacity breakdown — protected and orphaned volume counts with provisioned and physical bytes, searchable and paginated — plus a per-profile drill-down listing every bound volume with its origin and drift.
• Allowed rebinding an adopted relationship to a different automation profile and unbinding it back to adopted; unbind is refused while an autonomous grow or approved cleanup is still in flight.
• Added a free-text operator note on each relationship.
• Recorded per-cluster TLS posture (verification-skip and trust-anchor fingerprint, never raw certificates or credentials) in the audit trail.
• Hardened notification coalescing with a database-safe conditional insert, and encoded ONTAP request query parameters.

1.0.0 — initial stable release

• Introduced volume-level SnapMirror observation and destination-side orchestration for customer-managed ONTAP and Amazon FSx for ONTAP.
• Added policy matching, human approvals, dry-run safety, notifications, and audit history.
• Added durable PostgreSQL checkpoints for destination volume creation, relationship creation, initialization, remediation, and orphan cleanup.
• Added bounded fill-and-grow behavior and approval-only orphan deletion.
• Added multi-replica reconciliation guarded by a PostgreSQL advisory lock.

Maintaining this changelog

1. Record completed operator-visible controller work under the target release's version heading.
2. Use the immutable application image version, and add the date when the release is published.
3. Link deployment instructions to that immutable application image tag.
4. Keep published release entries unchanged; document corrections in a newer release.