TAK-OS Replicated CSI documentation

TAK-OS Replicated CSI provides coordinated RWX and RWO filesystem volumes backed by local-CSI PVCs on multiple Kubernetes nodes. An elected replica coordinator orders operations; Kubernetes manages lifecycle without carrying ordinary file I/O. The driver does not discover or format raw disks.

Current release: 4.1.1, updating user and administrator documentation for the shared NODE runtime introduced in 4.1.0. Shared mode remains for staging evaluation and retains the credential-renewal fix under the scope introduced in 4.0.0. Release images build and publish through TLSPU CI. Historical runtime observations have their own source/image boundaries; they are not verification of a new floating dependency build.

Release 4.1.0 contains a shared NODE runtime that replaces generated controller/worker Pods. It requires the companion local CSI consumer extension in 2.1.0 and reviewed CRD/NODE/RBAC/mount configuration. See architecture and the transition procedure; this source has compilation evidence only and has not been rolled out or observed on a cluster.

User documentation

Start here if you run an application on an installed cluster.

• User guide — choose RWX/RWO, request and mount a claim, initialize permissions, publish files, diagnose errors and handle backups or claim removal.
• Supported capabilities and limits — suitable workloads, file operations, failure tolerance and excluded features.
• Filesystem semantics — visibility, durability, application coordination and implementation ceilings.

Administrator documentation

Start here if you install the driver or operate storage for application teams.

• Administrator guide — deployment planning, storage policy, inventory, monitoring, repair, maintenance, backup and retirement.
• Getting started — prerequisites, GitOps installation order and StorageClass/PVC examples.
• Configuration reference — CLI options, shared NODE setup, cluster DNS, credentials, ports and resource settings.
• Operator runbook — incident diagnosis and guarded repair procedures for existing volumes.

The sections below provide detailed references and historical records for both audiences. Installation and maintenance examples require an explicitly selected, authorized cluster; reading these guides does not perform a rollout.

Architecture and filesystem behavior

Learn where authority lives and what an acknowledged operation means.

• Architecture — roles, control/data paths and identity model.
• Replication and durability — quorum, ordering, retry outcomes, checkpoint transfer, restart and controlled replacement.
• Filesystem semantics — permissions, visibility, rename/append, sparse files, links and unsupported operations.
• CSI and FUSE sessions — stage/publish, split frontend/mount owner, path validation and unmount recovery.
• RWO ownership — node leases, epoch fencing and takeover.
• Durable handles and links — open-unlink lifetime, renewal/expiry and final-reference reclamation.
• Volume lifecycle — provisioning, immutable ownership, deletion, Retain and safe uninstall.

Operations and recovery

Operate existing volumes without substituting API object deletion for storage fencing or discarding original identity.

• Operator runbook — diagnostics, repair requests, credential renewal, degraded mounts and cleanup blockers.
• Observability — logs, metrics, readiness and interpretation of journal/checkpoint/rebuild progress.
• Capacity and performance — actual storage consumption, implementation bounds and historical measurements.
• Security and resource bounds — trust model, authentication, credential overlap and bounded work.
• Backup and disaster recovery — external archive tools, application quiescence and restoration into new authority.
• Upgrades and compatibility — quiescent role updates, preserved identities and unsupported mixed-version/downgrade operation.

Builds and releases

Use the established shared CI process and publish explicit release notes.

• Builds and OCI images — per-role Dockerfiles, floating inputs, image contents and template-only CI jobs.
• Release policy and procedure — changelog-before-tag rule, versioning, release publication and operator authorization.
• Changelog — versioned release history and unreleased changes.

Historical records

These pages describe earlier milestones rather than current installation or support requirements. Removed test tools are not pending operator tasks.

• 3.0 POC scope — original rewrite milestone and its limits.
• 3.1 implementation milestone — file/session and coordinated-recovery additions.
• 3.2 staging observations — recorded environment and source-bound operating observations.

For current instructions, use the guides above. For source files and archived reports, follow their explicitly labelled source/history links. Private backups, credentials and workstation evidence are not part of this documentation site.