docs/netdata-ai/skills/query-netdata-cloud/how-tos/INDEX.md
This directory holds operational how-tos: short, focused recipes that combine the per-domain guides into answers for specific questions. Each how-to documents the question, the steps taken, the wrappers used, and the expected output shape.
The how-tos catalog is meant to be live. Every time an AI assistant (or human) is asked a question that:
the assistant MUST author a new how-to in this directory and add it to the index BELOW before completing the task.
This is mandatory. Skipping it means the next assistant repeats the same analysis from scratch.
Filename: <slug>.md (e.g. find-node-id-by-hostname.md).
Sections:
query-netdata-agents/scripts/_lib.sh.Every code example must use the token-safe wrappers
(agents_query_cloud, agents_query_agent,
agents_call_function). No raw curl with -H "Authorization: Bearer $TOKEN" -- that defeats the no-token-leak guarantee.
(Populate as how-tos are authored. Stubs below mirror the canonical
skill-verification harness questions for verify/questions.md; replace each
(stub -- not yet authored) with a real link as soon as a how-to is written.)
find-node-id-by-hostname.md (stub -- not yet authored)find-node-hardware-specs.md (stub -- not yet authored)find-node-os.md (stub -- not yet authored)compare-explicit-and-room-wide-node-scope.md -- compare a fixed UUID scope with the current room-wide node scope; explains why all-room queries omit scope.nodes and use selectors.nodes: ["*"], why a large UUID selector is redundant and expensive, and how to pass large payloads through the token-safe wrapper with @file.fleet-connectivity-slo-queries.md -- single-dimension fleet percentages (percent of devices connected/streaming, percent of devices with a boolean dimension at 1 or 0) and ranking devices by percent of time a boolean dimension was 0; includes the average-of-boolean trick and the countif-through-Cloud caveat.is-node-a-parent-and-children.md (stub -- not yet authored)is-node-a-child-and-parent-target.md (stub -- not yet authored)list-vnodes-on-node.md (stub -- not yet authored)diagnose-no-data-on-zoom-parent-retention-gaps.md -- why a node shows "No data" when zooming in while wider zoom renders fine: identify the serving agent from jsonwrap .agents, compare forced-tier queries (tier 0 vs 1 vs 2), reduce all-null rows to gap runs, run the decisive control test (does the PARENT's own local data have the same tier0 hole?), read the parent's daemon log via the windows-events/systemd-journal Function for DBENGINE write errors, and quantify child streaming flapping via netdata.streaming_outbound replicating buckets.find-failed-collection-jobs.md (stub -- not yet authored)is-collector-monitoring-X-and-frequency.md (stub -- not yet authored)pid-with-biggest-memory-and-app-group.md (stub -- not yet authored)currently-firing-alerts-in-room.md (stub -- not yet authored)alert-config-by-cfg-hash.md (stub -- not yet authored)silenced-alerts.md (stub -- not yet authored)last-netdata-status-file-log.md (stub -- not yet authored)recent-error-logs-in-namespace.md (stub -- not yet authored)local-l2-topology-summary.md (stub -- not yet authored)group-network-topology-by-kubernetes-pod.md -- summarize topology:network-connections process actors by Kubernetes pod and namespace through Cloud.find-containers-for-topology-port.md -- find containers or pods exposing a specific TCP port from the Cloud topology Function payload.validate-local-netflow-function.mdtop-flow-talkers-last-hour.md (stub -- not yet authored)members-by-role-in-space.md (stub -- not yet authored)rooms-with-most-nodes.md (stub -- not yet authored)node-state-changes-last-hour.md (stub -- not yet authored)When the answer needs both Cloud-side and direct-agent-side calls (e.g. "find the parent of a stale node, then read its streaming-state directly"), author the how-to under the skill that owns the FIRST wrapper call and cross-link to the other.