Reliability defaults
Fair resource sharing is now enabled by default in charted deployments so new installs and upgrades avoid accidentally disabling workload scheduling fairness.On-premises 10.x installs
Metoro Hub 10.x now has a customer-facing standalone Helm chart for on-premises deployments. The chart uses an on-prem-specific values API, validates common configuration mistakes before install, and supports bundled or external dependencies without exposing SaaS-only chart options.- Bundled PostgreSQL now runs on CloudNativePG with operator-managed failover and optional scheduled object-store backups.
- Bundled ClickHouse has production-oriented defaults for startup behavior, disruption protection, Keeper metrics, hot/cold object storage, and optional full backups.
- Monitored-cluster onboarding can install the pinned on-prem exporter chart directly, including support for ServiceMonitor-based baseline metrics.
- Trusted CA bundles can be passed through the Hub and exporter charts for private PKI, self-signed endpoints, PostgreSQL TLS, and object-store backups.
Chart inspection improvements
Chart tooltips are easier to inspect during debugging. You can pin a tooltip by clicking the chart, keep crosshairs aligned across related charts, drag pinned tooltips, filter long tooltip series lists, and distinguish display labels from split/group-by labels more clearly.Trace debugging
Service APM trace feeds now include status controls that make it easier to focus on failures. You can filter traces by available HTTP status codes or OpenTelemetrystatusCode values, and quickly toggle to error-only traces from the service trace view.AI investigation from traces
Error trace details now include a Guardian investigation action that sends the trace context to AI SRE so you can ask what happened from the trace view. Related logs also use the server container and the trace time range, so useful context can appear even when logs do not explicitly contain a trace ID.Service map context
Service map edges now show compact trace-count badges directly on the map, giving connection-volume context without opening the details panel.Faster repeated historical queries
Metoro now has an incremental query response cache for repeated historical dashboard and explorer queries. Stable long-range metric, log, trace, and Kubernetes queries can reuse previously computed work, making repeat views of historical windows significantly faster once cached.Log listing also tries a short edge window first when loading the first page of recent logs, then falls back to the full range only when needed. This makes dense log views feel more responsive without changing export behavior.AI SRE conversation history
Guardian conversations now get concise chat summaries so new chats are easier to recognize later. Autonomous investigation runs are hidden from normal chat history, keeping “My conversations” focused on user-started chats while preserving automation outputs in their run details.Public alert automation
Public alert definitions can opt into AI investigations when alerts fire, including alerts created through public API clients and external automation workflows.Faster Guardian execution
Guardian has a new execution runtime for chat and automated investigations. It improves benchmark results by 2x and makes long-running investigations more resilient when sessions reconnect or work resumes.Live automation progress
Deployment verification, alert investigations, anomaly investigations, service error investigations, issue-fixer automation, and repository auto-mapping now benefit from the same faster execution model. Active agent runs show live progress while work is running instead of waiting for final output.Public dashboard sharing
Admins can now share read-only dashboards outside their Metoro organization with password-protected public access. Public viewers unlock a shared URL with a dashboard password and receive a short-lived dashboard token, while Metoro limits public data access to admin-approved query shapes.Query allowlists
Public dashboard settings now extract widget queries for review, preserve approvals for unchanged queries, and mark new or changed queries as pending. Public metric, log, and trace endpoints reject requests that do not match an approved query shape.Viewer reliability
Public dashboard routes have been hardened so shared dashboards render with the same tooltip context as authenticated dashboards.Alert creation from charts
You can now create alerts directly from MetoroQL-backed chart menus. The alert creation flow starts with the chart query prefilled, matching the existing workflow for standard metric charts.Chart display controls
Chart Builder and Editor now include first-class threshold controls. Add, edit, remove, label, and color threshold lines directly from Display Options without editing dashboard JSON.Dashboard workflow polish
Dashboard charts can be duplicated within their original group, chart loading states now show only a spinner until data is ready, and grouped dashboard headers have clearer expand/collapse controls.More actionable CPU throttling issues
CPU throttling Advisor recommendations now identify the top contributing container when container-level metrics are available, including throttled seconds, CPU usage, and throttle ratio in the issue details drawer.SLO dashboard templates
Dashboard templates now include service-level objective templates, giving teams a faster way to start tracking availability, latency, and error budget health in Metoro.More dashboard widgets
Dashboards now support more Grafana-style panel types natively:- Table widgets for tabular metric data, with pagination, density, wrapping, column alignment, filtering, widths, precision, and unit controls.
- Bar Gauge widgets for comparing aggregate values as proportional bars.
- Pie and donut widgets for comparing aggregate metric values by category.
- Widget descriptions that appear as compact info tooltips and are preserved from imported Grafana panels.
Better Grafana and Perses imports
Grafana imports now preserve more dashboard behavior when moving into Metoro. Imported dashboards keep dense 24-column layouts, static interval variables, hidden variables, table units, bar gauges, pie charts, descriptions, and more MetoroQL compatibility for PromQL-style expressions such asround(...), time(), clamp_min(...), scalar comparisons, and histogram quantiles.