robintra/perf-sentinel chart-v0.2.61

perf-sentinel chart v0.2.61

on GitHub

What's new in chart-v0.2.61

Like chart-v0.2.57 and chart-v0.2.59, this is a real chart-level template change, and it also advances appVersion to track the new daemon release. It folds two chart versions into one tag: 0.2.60 ships two opt-in observability resources, and 0.2.61 moves appVersion from 0.8.12 to 0.8.13.

The chart now ships operator-facing alerting and disruption protection, both opt-in and off by default, so an upgrade that does not flip the flags renders an identical manifest set.

Added

• Opt-in PrometheusRule (prometheusRule.enabled). Packages the daemon's loss and saturation alerts so they are delivered out of the box instead of being a build-it-yourself wiring exercise. The perf-sentinel.rules group covers the daemon being unreachable (absent(perf_sentinel_active_traces)), OTLP rejection, analysis shedding, analysis-queue saturation, the findings store nearing its cap, correlator-pair eviction, and service-cardinality overflow, each alert naming the [daemon] knob to raise. Per-backend energy-scraper staleness alerts are gated behind prometheusRule.energyScrapers (off by default, only meaningful when an energy backend is configured), and prometheusRule.additionalRules appends custom rules verbatim.
• Opt-in PodDisruptionBudget (podDisruptionBudget.enabled). Adds voluntary-disruption protection during node drains and cluster upgrades. The default is maxUnavailable: 1 rather than minAvailable: 1, since the daemon runs single-replica and a minAvailable: 1 PDB would block every drain and wedge node maintenance. Set minAvailable only for a trace-aware sharded topology.

Changed

• appVersion bumped from 0.8.12 to 0.8.13, and the artifacthub.io/images tag follows to ghcr.io/robintra/perf-sentinel:0.8.13. The 0.8.13 daemon emits the SCI per-functional-unit intensity (co2.sci_per_trace) alongside the footprint, maps detectors to RGESN 2024 criteria, and adds an interpretive ESRS E1 crosswalk to the periodic disclosure (schema perf-sentinel-report/v1.3, additive). See the v0.8.13 release.

Behavior

• Default deployments are unaffected. Both new resources are off by default, so an upgrade renders the same manifest set as chart-v0.2.59 unless the operator sets prometheusRule.enabled or podDisruptionBudget.enabled.
• The appVersion bump rolls the pod to the 0.8.13 image. An upgrade that keeps the default empty image.tag resolves the image to ghcr.io/robintra/perf-sentinel:0.8.13 and reschedules the pod. The 0.8.13 changes are additive on every daemon and machine surface, no /api/* route, OTLP wire shape, configuration key, or Prometheus metric is removed or changed.
• The PrometheusRule needs the Prometheus Operator CRD. It renders the monitoring.coreos.com/v1 kind, so enable it only on a cluster that runs the operator (for example the kube-prometheus-stack). The PodDisruptionBudget uses the core policy/v1 API and needs no CRD.

Install

The chart is published as an OCI artifact on GHCR, install it directly with no helm repo add step:

helm install perf-sentinel oci://ghcr.io/robintra/charts/perf-sentinel --version 0.2.61

Upgrade an existing release:

helm upgrade perf-sentinel oci://ghcr.io/robintra/charts/perf-sentinel --version 0.2.61

Full Changelog: chart-v0.2.59...chart-v0.2.61