ktop Release v0.6.0

ktop v0.6.0 introduces Kubernetes Pressure Stall Information (PSI) resource contention metrics, a dedicated user session directory with rotated logging (~/.ktop/), dynamic in-app table column selection and configuration persistence (~/.ktop/config.yaml), interactive column sorting and filtering across subpages, major Prometheus scraper and metric store performance optimizations, codebase modernization, and toolchain upgrade to Go 1.27.

PSI (Pressure Stall Information) Metrics

• Resource Contention Signals: Integrated cAdvisor/kubelet PSI metrics (container_pressure_cpu_waiting_seconds_total, container_pressure_memory_waiting_seconds_total, container_pressure_io_waiting_seconds_total) across node and pod workloads.
• Overview STALL Column: Added a STALL column to Nodes and Pods overview tables rendering the dominant waiting stall percentage (CPU, Memory, or I/O) over the scrape window.
• Cluster Summary Stall Gauge: Added cluster-level PSI stall metrics to the Cluster Summary header.
• Version Guardrails: Added runtime version checks that surface ⚠ PSI beta in the cluster header when Kubernetes is <1.36 and prefix STALL values with ⚠ 0% when node kernels are <4.20.

User Session Directory & Rotated Logging (~/.ktop/)

• Session Directory: Introduced automatic creation of ~/.ktop/ (mode 0700) for user-specific data, relocatable via the $KTOP_DIR environment variable.
• Rotated Structured Logging: Integrated log/slog and lumberjack.v2 writing to ~/.ktop/ktop.log (mode 0600) with automatic 10MB rotation, retaining the last 3 sessions without interfering with the terminal display.
• CLI Logging Flags: Added --log-level (debug, info, warn, error) and --log-format (text, json) flags.
• Graceful Degradation: Application continues running cleanly if directory or file creation fails due to restrictive filesystem permissions.

Dynamic Column Configuration & Persistence (~/.ktop/config.yaml)

• Interactive Column Selector Modal: Added interactive column selection dialog triggered via the = hotkey on focused overview tables, supporting Space toggle, Enter apply, and Esc cancel.
• Active Default Columns: Enforced compact default column layouts for Nodes (NAME, STATUS, RST, PODS, DISK, CPU, MEM) and Pods (NAMESPACE, POD, READY, STATUS, RST, AGE, CPU, MEMORY), preventing horizontal table sprawling on standard terminals unless explicitly requested with --show-all-columns.
• Configuration Schema: Introduced ~/.ktop/config.yaml supporting table column customizations with resolved defaults for metrics collection and logging.
• Automatic Persistence: Column selections passed via CLI flags (--node-columns, --pod-columns) or modified via the interactive modal automatically persist to ~/.ktop/config.yaml.
• Modal Event Isolation: Isolated modal input capture so dismissing the column dialog with Esc restores table focus without triggering application-level exit.

Subpage Navigation, Sorting & UI Consolidation

• Shared Table Controller: Extracted reusable ui.TableController encapsulating sort toggling, shortcut key binding, slash / filter lifecycle, and scroll titles.
• Node Detail Pods Sorting & Filtering: Enabled interactive column sorting (c CPU, m MEM, n NAMESPACE, p POD, s STATUS, r RESTARTS, d READY, a AGE) and / slash filtering on the Node Detail Pods table.
• Header Shortcut Collision Fix: Renamed Node Detail Pods second column from NAME to POD, matching the Overview page and fixing the hotkey collision on 'N' with NAMESPACE.
• Format Consolidation: Unified detail formatting utilities (FormatCPU, FormatMemory, FormatDuration, FormatEventAge, Truncate, AdaptiveRateMax, CycleFocus) into ui/format.go.

Performance & Scraper Optimizations

• Scraper Loop Deduplication: Eliminated duplicate background scrape goroutines in KubernetesScraper, halving proxy scrape traffic to node kubelets and cAdvisor endpoints.
• O(S) Pod Metrics Aggregation: Rewrote GetAllPodMetrics from an $O(N_{\text{ns}} \times N_{\text{pods}} \times N_{\text{series}})$ Cartesian scan into a single-pass $O(S)$ aggregation grouped by (namespace, pod, container).
• Node Metric Query Deduplication: Reused existing node metrics during overview and summary refresh cycles to avoid duplicate queries.
• Pod-to-Node Indexing: Pre-indexed pods by Spec.NodeName in $O(P)$ when constructing node models.
• Sparkline Memory Leak Fix: Added automatic eviction of deleted pods and nodes from memory sparkline maps.
• InitContainer Metric Correctness: Corrected CPU request accounting in GetPodContainerSummary for pods configured with init containers.
• Informer Resilience & RBAC Fallback: Replaced background informer sync panics with graceful error recovery, supporting namespace-scoped RBAC clusters.
• TUI Concurrency Deadlock Fix: Switched refresh queue delivery to non-blocking sends and debounced filter inputs to eliminate UI freezes.

Codebase Simplification & Toolchain Modernization

• Dead Code Removal: Removed unused views/pod/logs.go and deprecated metrics informer/lister infrastructure in k8s/.
• Direct Metric Summation: Eliminated redundant v1beta1.PodMetrics round-trip conversion in MainPanel.refreshPods.
• Go 1.27 Toolchain: Upgraded module toolchain to Go 1.27 across go.mod, Dockerfile, and GitHub Actions workflows.
• Standard Library Idioms: Modernized codebase to use slices.Sort, slices.Contains, slices.Clone, and standardized error wrapping with %w.
• Multi-OS CI Matrix: Expanded GitHub Actions CI test suite to validate builds and tests across Linux, macOS, and Windows (windows-latest).