Skip to main content

Observability

Pixee Enterprise ships a complete observability stack -- metrics, logs, traces, and dashboards -- bundled in the Helm chart. No separate purchase or manual setup required. SRE teams can use the bundled stack as-is, disable individual components, or route telemetry to their existing observability platform. This page covers what is included, how to integrate with your existing stack, and key metrics to monitor.

Bundled Observability Stack

The Helm chart includes five observability components, each shipped as a conditional subchart that can be independently enabled or disabled:

ComponentPurposeReplaceable?
VictoriaMetricsMetrics collection and storageYes -- BYO Prometheus or VictoriaMetrics
Victoria-logs + collectorLog aggregationYes -- BYO log pipeline
Victoria-tracesDistributed tracingYes -- BYO tracing solution
Grafana k8s-monitoringDashboards and visualizationYes -- BYO Grafana
SentryError reportingToggleable (opt-in / opt-out)

Pixee ships curated dashboards built specifically for the Pixee platform -- not just upstream chart defaults. These dashboards cover platform health, analysis throughput, LLM performance, and fix quality metrics.

Why this matters for enterprise evaluation: Legacy on-premises scanners (Checkmarx, Veracode) predate Kubernetes-native observability and do not ship monitoring. SaaS scanners do not expose observability because the customer does not run the infrastructure. Pixee ships both the platform and the tools to monitor it.

Integrating with Your Existing Stack

Most enterprise platform teams already run an observability stack. Pixee integrates rather than duplicating.

Disable embedded subcharts. Turn off any embedded component (VictoriaMetrics, Victoria-logs, Victoria-traces, Grafana) and configure Pixee to emit telemetry to your existing platform. Each subchart is independently toggleable.

Grafana Cloud integration. Grafana Cloud Private Data Connector support routes metrics from your self-hosted Pixee deployment to your Grafana Cloud tenant without exposing the cluster to inbound connections.

Standard pipelines. Pixee's telemetry is compatible with standard Prometheus and OpenTelemetry pipelines. If your team already runs Prometheus, Datadog, or a similar platform, Pixee's metrics integrate without custom adapters.

Verification. After configuring your integration, verify that metrics, logs, and traces are flowing to your platform. Pixee's admin console shows telemetry status for embedded components.

Key Metrics to Monitor

For day-two operations teams, these are the metrics that indicate platform health and business value:

Metric CategoryWhat to WatchWhy It Matters
Analysis throughputTasks completed per hour, queue depthPlatform is processing findings at expected rate
LLM latencyInference response time per tierPerformance tuning; identifies slow or rate-limited providers
Fix qualityAverage safety, effectiveness, and cleanliness scoresQuality monitoring; trends indicate model or configuration changes
Merge ratePercentage of PRs merged by developersAdoption tracking; the primary business-value metric
Error rateFailed analyses, retries, suppressed fixesOperational health; spikes indicate configuration or LLM issues

Business metrics vs. platform metrics: Merge rate, triage volume, and remediation velocity are business metrics visible in Pixee's reporting dashboard. Analysis throughput, LLM latency, and error rates are platform metrics visible in the observability stack. Both matter for different audiences -- share business metrics with security leadership and platform metrics with SRE.

For business-level reporting (merge rate trends, triage reduction, MTTR), see Enterprise Overview > Measuring Success.

Operational Controls

Operational settings are configurable through the KOTS admin console (embedded cluster) or Helm values (BYO Kubernetes):

ControlDescription
Sentry toggleEnable or disable error reporting
Custom metrics toggleEnable or disable custom Pixee metrics
LLM debug modeVerbose logging for LLM request/response troubleshooting
Support bundle configurationLog size and age limits for diagnostic bundle generation

For troubleshooting operational issues, see Enterprise Troubleshooting.