v5

Highlights

• Deployment streams now stay accurate after deletions by clearing removed Flink deployments from the watcher cache.

Improvements

• Prevented deleted deployments from reappearing for clients after reconnects or cached watcher state replay.

API cleanup

• No API changes in this release.

Internal changes

• Tightened Kubernetes delete-event handling in the deployment watcher cache.

v4

Changes in this release

Deployment Recovery

• Added support for recovering deployments from stored state.
• Extended deployment state handling in the backend and frontend so recovery actions are available from the deployment UI.

Storage Checkpoints and Savepoints

• Added storage checkpoint metadata inspection, including a dedicated detail view for stored checkpoint/savepoint entries.
• Unified storage checkpoint and savepoint listing so both entry types are presented through a consistent backend API and frontend table.
• Improved S3-backed checkpoint/savepoint discovery and metadata loading.

Frontend Streaming

• Extracted the shared SSE streaming logic into the reusable useSseStream hook.
• Simplified deployment stream handling on top of the shared hook.

Backend Maintenance

• Applied backend lint cleanup following the iferrinline pattern.
• Added custom golangci-lint configuration and updated CI to run the custom lint wrapper.

Full Changelog: v3...v4

v3

Highlights

• Add suspend and resume controls for Flink deployments in both the deployments table and deployment detail pages.
• Introduce admin mode gating so deployment actions stay hidden until temporary admin access is enabled from the UI.
• Add tabbed deployment views for All and Not running deployments to speed up triage.

Improvements

• Preserve the selected deployment view when navigating between the list and deployment details.
• Let the Not running view surface issues across namespaces for faster operational investigation.
• Highlight suspended deployments more clearly and normalize deployment detail timestamps to UTC for easier reading.

API cleanup

• Add backend suspend and resume endpoints for Flink deployments.
• Reuse shared deployment selector handling across related backend endpoints for more consistent behavior.
• No breaking API removals are included in this release.

Internal changes

• Refactor frontend context, hooks, and formatting utilities to support admin mode and deployment lifecycle actions.
• Remove outdated workflow notes while aligning the UI and backend plumbing around the new release features.

v2 - Flink Admin

A maintenance-focused release that modernizes the runtime setup and makes local development workflows easier to run and repeat.

Highlights

• Upgraded the backend to gosoline v0.57.2 and refreshed the Go dependency set to stay aligned with the current framework stack.
• Migrated backend configuration to the newer app structure used by recent gosoline versions.
• Added mise tasks for frontend and backend build, lint, test, and local CI workflows.
• Refined the dashboard issue filter so finished jobs no longer show up as active problems in the "Not Running" view.

Upgrade Note

• Custom backend config files should move from the legacy top-level env and app_* keys to the new nested app block.
• Embedded frontend asset handling was also cleaned up, so generated files are no longer tracked in git.

Included In This Release

• Dependency and configuration modernization
• Repeatable local build and CI commands
• A small dashboard quality-of-life fix for operations views

v1 - Flink Admin

A web-based administration interface for Apache Flink clusters running on Kubernetes. Monitor deployments in real time, inspect checkpoints and savepoints, browse Kubernetes events, and diagnose job failures — all from a single UI backed by a single self-contained binary.

---

Deployment Dashboard

• Live table of all FlinkDeployment resources across all namespaces, updated in real time via Server-Sent Events
• Filter by namespace and lifecycle state; filters persist in the URL (bookmarkable/shareable)
• "Not Running" quick-filter highlights deployments with a non-running job state
• Flink UI deep-link icon on each deployment row when a Flink web UI is reachable
• Connection status indicator with automatic reconnection and manual retry
• Columns: Name, Namespace, Lifecycle State, Job State, Flink Version, Image tag, Parallelism, JM/TM resource allocations, Age

Deployment Detail View

• Status summary header: Lifecycle State, Job State, JobManager Deployment Status, Age, Job Start/Update times
• Five tabs for each deployment — Details, Checkpoints & Savepoints, Storage, Events, Exceptions
• Badge counts on the Exceptions and Events tabs so issues are visible at a glance without opening the tab
• Back button returns to the dashboard with your previous filters intact

Deployment Details Tab

• Spec overview: image, Flink version, parallelism, entry class, JAR URI, upgrade mode
• Job arguments rendered as a readable key/value table
• JobManager and TaskManager resource allocations (CPU, memory, replicas)

Checkpoints & Savepoints Tab

• Fetches live data from the Flink REST API (requires an ingress on the deployment)
• Summary counts: Total, Completed, Failed, In Progress, Restored
• Checkpoint history table: ID, status (color-coded), type, trigger time, duration, state size, subtask acknowledgement progress, storage path

Storage Tab

• Lists checkpoints and savepoints directly from AWS S3 — no Flink REST API required
• Validates checkpoint directories by checking for a _metadata file in each chk-* folder
• Shows checkpoint name, Job ID, last modified time, size, and full S3 path
• Savepoints are scoped to the currently active Job ID
• Configuration card shows the checkpoint and savepoint directories from the deployment's Flink configuration

Kubernetes Events Tab

• Lists events.k8s.io/v1 events for the deployment, sorted newest-first
• Columns: Age, Type (Warning / Normal, color-coded), Reason, Message, Source, Action

Exceptions Tab

• Fetches the exception history from the Flink REST API (requires an ingress on the deployment)
• Table: timestamp, exception name, task name, location, failure labels
• Expandable rows with full stack traces and nested concurrent exception details
• Truncation notice when more than 50 exceptions exist

---

Configuration & Deployment

• Ships as a single static binary with the React frontend embedded — no separate web server needed
• Docker image published to ghcr.io for linux/amd64 and linux/arm64
• Kubernetes authentication via in-cluster service account (default) or kubeconfig for local development
• S3 region defaults to eu-central-1; fully configurable via environment/YAML config
• HTTP server runs on port 8082; the frontend dev server proxies /api requests automatically

Requirements

• Kubernetes cluster running the Flink Kubernetes Operator (flink.apache.org/v1beta1)
• AWS S3 access with read permissions on the configured checkpoint/savepoint buckets (Storage tab)
• Ingress configured on FlinkDeployments for Flink REST API features (Checkpoints & Savepoints tab, Exceptions tab)