perf-sentinel chart v0.2.62

What's new in chart-v0.2.62

A small chart-only patch. It drops one noisy alert from the PrometheusRule shipped in chart-v0.2.60. appVersion stays at 0.8.13, the daemon image is unchanged, and no other template changes.

Removed

• Dropped the PerfSentinelFindingsStoreNearCap alert. The findings store is a bounded ring buffer that evicts the oldest entry by design, and perf_sentinel_findings_total keeps the authoritative cumulative count, so reaching the cap is normal operation rather than data loss. On a busy daemon the gauge pins at its cap and the alert fired continuously, which is noise, not a signal. The eviction alert that does matter, PerfSentinelCorrelatorEvicting (cross-trace correlations actually dropped under load), is kept.

Behavior

• Only affects clusters that enabled the rule. The PrometheusRule is opt-in (prometheusRule.enabled, off by default), so a deployment that never enabled it renders an identical manifest set. The remaining alerts are unchanged: daemon unreachable, OTLP rejection, analysis shedding, analysis-queue saturation, correlator-pair eviction, service-cardinality overflow, and the opt-in per-backend energy-scraper staleness alerts.
• No daemon change. appVersion stays 0.8.13, so an upgrade does not roll the pod for an image change. The only difference is one fewer alert in the rendered PrometheusRule when prometheusRule.enabled is set.

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 0.2.62

Upgrade an existing release:

helm upgrade perf-sentinel oci://ghcr.io/robintra/charts/perf-sentinel --version 0.2.62

Full Changelog: chart-v0.2.61...chart-v0.2.62