Skip to content

v0.5.2 — MCP metrics and Grafana dashboards

Choose a tag to compare

@alexconrey alexconrey released this 07 Aug 19:37
· 140 commits to main since this release
5d257de

What's new

MCP observability

  • deckwatch_mcp_requests_total{method,status} + duration histogram on every JSON-RPC call through /mcp
  • deckwatch_mcp_tool_calls_total{tool,status} + duration histogram for deckwatch-native tools (create_application, set_gitops, trigger_build, plugins, monitoring, etc.)
  • K8s-passthrough tools are self-instrumented by the mcp-k8s library via mcp_k8s_tool_calls_total — no double-counting

Grafana dashboards

Two pre-built dashboards added under dashboards/:

  • Deckwatch Overview — HTTP traffic, GitOps pipeline, k8s API, frontend RUM (Core Web Vitals), and an AI/MCP activity row
  • Deckwatch MCP — deep-dive MCP dashboard with per-method request rates, per-tool call rates, deckwatch-native vs k8s source split, p95/p99 latency, permission denials, and error breakdown

Dependency updates

  • mcp-k8s bumped to 89f584a — includes library-level tool metrics (PR #7) and new update_deployment_labels / update_deployment_annotations / update_deployment_affinity tools (PR #9)
  • Helm chart version bumped to 0.5.2 (lockstep with app version)