Lock dashboards and see runtime metrics on spans

Improvements Fixes

TL;DR

Authors and admins can lock a dashboard from its menu, long dashboards get a section jump menu, and span details show runtime and host metrics.

Lock dashboards and see runtime metrics on spans

Improvements

  • AI Assistant: Ask for a service’s request count and 5xx count in one question, and the assistant reads both instead of asking which measurement you meant. When you name a service and discovery finds exactly one match, it uses that service instead of asking you to pick.
  • Dashboards
    • Lock or unlock a dashboard with Lock and Unlock, after Edit in the dashboard menu and in the row menu of the dashboards list. Only the author or an admin can use them; other editors see the item disabled. A locked dashboard shows a lock icon after its name, and its tooltip names who can unlock it.
    • Added Duplicate to the dashboard page menu. It copies the saved version and opens the copy.
    • Added a jump menu to the dashboard header, labeled with the number of sections, that scrolls to a section and opens it. It replaces the Collapse all and Expand all row and the section filter. Picking a section adds it to the URL, so a shared link opens at that section.
    • Section headers and the variables bar are more compact, so panels start higher on the page.
    • A dashboard pinned to the Overview page shows every section heading with the first one open, and its header shows the time window. Resetting the zoom returns to the window the card opened with.
  • Traces
    • Added Runtime and Infrastructure tabs to span details. They show the runtime dashboard for the span’s process.runtime.name and the host or pod dashboard for its host.name or k8s.pod.name, over 15 minutes either side of the span, with its start and end marked.
    • Rows in the Traces Explorer span table fit on one line, with Service in its own column. A synthetic check span shows one chip in Type & Status, colored by the check result, that links to the check.
  • General: Discover Services, Applications (RUM), Databases, ECS, and Jobs, Synthetic Monitoring checks, the dashboards list, Changeboards, Profiles, and the Control Plane rule lists show an error with a Retry button when they fail to load. When filters hide every row, they say so and offer Clear Filters.

Fixes

  • Dashboards
    • Stat, gauge, and bar panels sorted series by value, so a pinned series such as a total moved when the time range changed. They now keep the order of the panel’s queries.
    • A stat panel with six or more values squeezed its tiles into the left half of the panel.
  • Discover Services
    • The Services breadcrumb on a service’s page built a malformed link that dropped the filters you had applied.
    • The Logs tab of a service showed no logs when the service’s logs carry the environment under a key other than deployment.environment, such as env.
    • Runtime dashboards under Related Metrics were empty when the metrics carry job but no service_name.
  • Logs: In Logs Explorer, the attribute breakdown popover scrolled sideways, hiding the Occurrence percentages and the filter buttons.
  • Coding Agents
    • The Claude Code setup snippet left out CLAUDE_CODE_ENABLE_TELEMETRY=1 and the exporter settings, so following it sent no telemetry.
    • Claude Code sessions could show no user email when the first event of a session did not carry one.
  • General: Two series on one chart could get nearly the same color.