All plan tasks

As of Nov 18, 2026 04:42 UTC. 52 tasks: 1 done · 16 proven in lab, in review · 6 in progress · 29 not started. Back to the summary.

Baselines and scorecard · 3 of 5 proven in the lab

TaskDeliverableStatusPasses whenLatest result
T0.1Public repo audit on `main`: `SUSE/elemental`, `SUSE/elemental-lifecycle-manager`, `rancher/elemental-operator`, `rancher/elemental-ui`, `rancher/elemental-channels`, `rancher/turtles`, CAPRKE2, Harvester repos; open PRs and issues that overlap this planDoneWritten findings; overlapping work linked to task IDs so nothing is built twiceNov 12: public repo audit; plan corrections and new blockers
T0.2`parity/matrix.yaml`: every user journey × baseline × target, with legacy step count and expected resultProven in lab, in reviewOwner approves the journey listNov 12: parity matrix draft (52 journeys) for owner review
T0.3Journey harness: scripted Rancher API + UI (Playwright) runs recording steps, wall time, pass/failProven in lab, in reviewRuns green against provisioning v2 on HarvesterNov 18: two clean Elemental 3 create samples on 3 + 2 (27m03s, 27m13s; legacy 25m30s)
T0.4Legacy baselines measured for all four baselinesIn progressNumbers committed, reproducible from one commandNov 14: Elemental 2 OS upgrade with 4 vCPU / 16 GiB control planes
T0.5Scorecard page generated nightly from harness resultsNot startedPublished; red/green per row

Images · 5 of 6 proven in the lab

TaskDeliverableStatusPasses whenLatest result
T1.1Output types beyond `iso` and `raw`: qcow2, VMDK/OVA, AMI, Azure VHD, GCE image, PXE artifactsProven in lab, in reviewEach boots on its target in CINov 12: Elemental 3 main boots on Harvester with boot-time config
T1.2Per-platform enablement: Ignition platform IDs, guest agents (qemu-guest-agent, open-vm-tools) as systemd extensionsNot startedNode reports IP and readiness to its platform
T1.3Prebuilt fast path: pipeline publishes signed tier 2 images per target for each Core Platform releaseProven in lab, in reviewA first cluster on Harvester needs no buildNov 16: prebuilt image pipeline; first cluster on Harvester with no build (lab)
T1.4Generic CAPI-ready image: `split` mode, no baked node role or network; all per-node config arrives at bootProven in lab, in reviewSame image used for server and agent nodesNov 16: generic CAPI-ready image; one image serves 3 control plane + 2 workers (lab)
T1.5Image build operator: CRD requesting (release, target, config) → `elemental3 customize` job → delivered to target; cache by digestProven in lab, in reviewSolution release requested in Rancher becomes a usable Harvester image with no manual step; second request is a cache hitNov 17: image build operator; request to usable Harvester image with no manual step (lab)
T1.6Unprivileged or contained buildsProven in lab, in reviewBuild job runs without host-level privileges, or in an isolated builder documented as suchNov 17: pod builder proof on the Rancher management cluster

Cluster provisioning · 4 of 6 proven in the lab

TaskDeliverableStatusPasses whenLatest result
T2.1Bootstrap contract: CAPRKE2 Ignition output consumed natively by Elemental 3; remove the manual Butane and kernel-parameter stepsProven in lab, in reviewCAPM3 cookbook example works with zero hand edits
T2.2ClusterClass per target (Harvester, Metal3, AWS, Azure, GCP, vSphere), parameterised by Elemental releaseProven in lab, in review`kubectl apply` of one Cluster object yields a ready clusterNov 18: Rancher-free create with a direct Harvester credential (17m46s, zero packets to Rancher)
T2.3Auto-import into Rancher with agents, projects, RBAC workingProven in lab, in reviewCluster shows Active in Rancher with no manual import
T2.4Manage operations mapped: scale, remediation (MachineHealthCheck), etcd snapshot/restore, cert rotationNot startedEach WS0 Manage journey passes
T2.5Compatibility with Turtles current and next minor (v1beta2 contract)Not startedCI matrix green on both
T2.6cloud-init support in Elemental 3 through yip, reusing elemental-toolkit work where possible; coexists with IgnitionProven in lab, in reviewA cloud-config supplied as NoCloud or ConfigDrive user-data configures an Elemental 3 node at first boot; CAPHV and Harvester node configuration work unmodifiedNov 12: cloud-init via yip design, prototype branch reference and lab proof

Upgrade · 2 of 6 proven in the lab

TaskDeliverableStatusPasses whenLatest result
T3.1Release catalog on the management cluster: manifest sources, discovered tier 2/3 releases, valid upgrade paths per clusterProven in lab, in reviewNew release appears without operator action; invalid jumps are not offered
T3.2Rancher-driven upgrade: LCM present in every Elemental 3 cluster by default; management controller sets the downstream `Release` and mirrors its statusIn progressPicking a release in the management cluster upgrades a 3+3 Harvester guest cluster with no `kubectl` on the downstreamNov 13: real version upgrade and power-cut injection; gaps G1-G13
T3.3LCM hardening: health-gated progress, concurrency control, fix early status.version, component config outside upgrades. Release rollback/downgrade out of scope (owner 2026-11-12)In progressInjected node failure stops the upgrade, the failed OS transaction rolls back via Elemental snapshots, and the cluster stays healthyNov 17: upgrade on a cluster Rancher does not manage (runs G and H)
T3.4Rolling-replace option for VM and cloud: CAPI rollout to an image of the new release (from T1.5)Proven in lab, in reviewZero-downtime upgrade by replacement; user chooses strategy per clusterNov 17: upgrade by replacement on 3 + 2 (54m39s); zero-downtime not met (API gaps up to 27 s)
T3.5Pre-flight checks and progress reporting (conditions, events, phase, percent) exposed for the UINot startedHarness reads progress through the management API only
T3.6API maturity: `Release` to v1beta1, manifest schema v1, conversion testedNot startedUpgrade across the schema change tested

Rancher UI · 1 of 5 proven in the lab

TaskDeliverableStatusPasses whenLatest result
T4.1UI extension: create-cluster wizard per target, matching provisioning v2 screens field for field where concepts overlapProven in lab, in reviewStep count ≤ provisioning v2 for the same clusterNov 18: Elemental 3 create-cluster wizard in Rancher; tenant created a real 1 + 1 (12 interactions vs 18 legacy)
T4.2Release picker and one-click upgrade with pre-flight result, progress, and rollbackIn progressUpgrade journey ≤ legacy steps; shows OS + RKE2 + apps as one versionNov 18: owner-approved owner binding for the UI test cluster (orchestrator shim log)
T4.3Bare-metal views: registration endpoints, inventory, select hosts into a clusterNot startedElemental 2 users complete the flow without docs
T4.4Cluster detail parity: machine pools, conditions, events, snapshotsIn progressAll Manage journeys doable from UI
T4.5Appears in Cluster Management next to existing provisioners, not in a separate areaNot startedUX review sign-off

Bare metal · 0 of 5 proven in the lab

TaskDeliverableStatusPasses whenLatest result
T5.1Registration service and inventory on Elemental 3: registration URL, seed/installer ISO, TPM-based identity, labelsNot startedHost boots ISO and appears in inventory
T5.2Inventory → CAPI infrastructure provider (hosts selected by label into clusters)Not startedCluster built from inventory hosts
T5.3Metal3 directed provisioning with Elemental 3 imagesNot startedEdge "directed network provisioning" journey passes
T5.4Disk encryption (#627), reset/re-provision, node decommissionNot startedJourneys pass
T5.5Edge Image Builder equivalence: document and automate the EIB → `elemental3 customize` mappingNot startedAn EIB definition converts and builds

Harvester · 2 of 8 proven in the lab

TaskDeliverableStatusPasses whenLatest result
T6.1Target: Harvester CAPI provider ClusterClass; image auto-upload to Harvester; Harvester cloud provider and CSI as solution chartsProven in lab, in reviewGuest cluster with load balancer and PVCs working
T6.2Tenancy: provisioning scoped by Rancher project and Harvester namespace, quotas respected, per-tenant credentialsNot startedTenant A cannot see or use tenant B resources; quota breach is refused cleanly
T6.3Parity with the Harvester node driver experience from Rancher (Virtualization Management → create cluster)Not startedStep count ≤ node driver path
T6.4Migration design spike: compare the Elemental 1 on-disk layout and configuration used by Harvester (partitions, active/passive images, cloud-init config) with Elemental 3 (btrfs snapshots, Ignition); choose an in-place conversion approachProven in lab, in reviewWritten design reviewed by the Harvester team; prototype converts one node in a VM with data intactNov 13: rolling three-node Harvester migration with live guest, rollback and Secure Boot
T6.5Harvester as a tier 3 solution manifest on SLES 16.1: host dependencies as systemd extensions, installer for new installsIn progressNew Harvester cluster installs from an Elemental 3 image and runs a VMNov 13: three-node rolling migration onto the host image with enforcing trial boots
T6.6In-place migration delivered through Harvester's existing upgrade flow: node by node, VMs live-migrate, persistent data and settings preserved, rollback if a node failsNot startedA 3-node Harvester cluster on the current Micro 6.x release reaches the Elemental 3 release from the normal upgrade UI, with running VMs and no reinstall
T6.7After migration, Harvester upgrades through LCM and the release catalog like any other solutionNot startedNext upgrade is a release pick
T6.8Harvester host operations on Elemental 3: add a node, put a node into maintenance, remove a node (journey J-MANAGE-11; owner ruling Q10, 2026-11-18)Not startedJ-MANAGE-11 passes with step count <= Harvester's own flow

Clouds and vSphere · 0 of 4 proven in the lab

TaskDeliverableStatusPasses whenLatest result
T7.1AWS: published AMIs, CAPA ClusterClass, cloud credentials reused from RancherNot startedDeploy/manage/upgrade journeys pass
T7.2Azure and GCP, same shapeNot startedSame
T7.3vSphere: OVA template publishing, CAPV ClusterClassNot startedSame
T7.4Cloud provider / CSI charts per platform in the solution manifestsNot startedLoad balancer and storage work

Air-gap, migration, docs · 0 of 3 proven in the lab

TaskDeliverableStatusPasses whenLatest result
T8.1Air-gap: mirrored registry bundle per release; all journeys with no internetNot startedHarness passes in an isolated network
T8.2Migration tooling: provisioning v2, Elemental 2, and Edge clusters → Elemental 3 (rolling node replacement into the new model)Not startedOne cluster of each type migrated with workloads up
T8.3Product docs, quick starts per target, and the parity report for stakeholdersNot startedDocs reviewed; a new user completes a quick start unaided

Quality gates · 0 of 4 proven in the lab

TaskDeliverableStatusPasses whenLatest result
T9.1Fix adoption blockers: SELinux with Cilium (#529), local-path (#480), HelmChartConfig (#570), manifest isolation (#516)Not startedIssues closed with regression tests
T9.2Nightly e2e matrix: target × release × journeyNot startedDashboard feeds the scorecard
T9.3Upgrade matrix: N-1 → N, skip attempts, mid-upgrade failuresNot startedAll handled or refused cleanly
T9.4Scale test: 100+ downstream clusters, 500+ inventory hostsNot startedNo degradation versus Elemental 2