ContextQMD
Back to Netdata

query-agent-events

.agents/skills/query-agent-events/SKILL.md

2.11.07.4 KB
Original Source

query-agent-events

Private developer skill for triaging crashes, panics, and fatals across the Netdata fleet. Reads the agent-events systemd-journal namespace via the Netdata systemd-journal Function (Cloud-proxied or direct-agent transport) and ships scripts that bake in index-friendly query patterns.

Why this skill exists

40k-200k status events arrive on the ingestion server every day on stable releases. The total fleet is 1.5M agents, so the dataset is large and noisy (many unupdated agents report crashes that have been fixed). Naive "grep all" queries are slow and wasteful. This skill teaches the maintainer (and any AI assistant helping them) how to slice the dataset efficiently and how to interpret what comes back.

Workflow

+-------------------------+        +---------------------+
|  get-events.sh          |  -->   |  <timestamp>.json   |
|  (cloud or agent API)   |        |  in .local/audits   |
+-------------------------+        +---------------------+
                                          |
                                          v
                              +------------------------+
                              |  analyze-events.sh     |
                              |  --by signal|version|  |
                              |       function|...     |
                              +------------------------+
                                          |
                                          v
                              +------------------------+
                              |  cluster + read source |
                              |  + fix the bug         |
                              +------------------------+

The skill is a bug-investigation tool, not a generic logs query tool. The two existing query-netdata-cloud and query-netdata-agents skills already cover transport mechanics; this skill EXTENDS them with the agent-events specifics (what fields are present, what predicates are index-friendly, what each enum value means for triage).

Key concepts (read first)

  1. The dataset: 40k-200k status events / day on stable releases, spread across 1.5M agents (not all restart daily). Naive full-namespace queries with bare FTS are slow.

  2. Index-friendly queries (HARD RULE): use multi-value field filters FIRST. The Netdata systemd-journal plugin supports the syntax:

    (FIELD1 in A, B, C) AND (FIELD2 in D, E, F) AND ...

    Between fields = AND. Between values = OR. This is a facet-engine feature, NOT raw journalctl. Use FTS via query= only as a residual narrower over the structured slice. See query-discipline.md.

  3. Three transports (priority order):

    • Cloud API -- proxied through Netdata Cloud at the agent-events space. Primary for the team.
    • Direct agent API -- against the agent-events node's /api/v3/function?function=systemd-journal. Primary for scripts.
    • ssh to the host -- operator-only path; mentioned in transports.md but no scripted ssh transport.

  4. After-the-fact event model: agents POST events ONLY on start (the previous session's exit reason). They commit status to disk on start, stop, and at most every 10 minutes. So the meaningful query unit is "events posted in the last 24 hours"; "the last hour" misses real crashes that haven't restarted yet.

  5. 23h client-side dedup (src/daemon/status-file-dedup.c:11): same agent + same event-content hash within 23h -> suppressed at the producer. So 1 record per agent per event-signature per day is the natural unit. Different agents posting the same crash signature -> both arrive (server does not dedup).

  6. Default time + version filters: 24h time window; latest stable + latest 2-3 nightlies for version. This focuses triage on bugs that still matter. Wide windows (--since '7d' or longer) are reserved for rare crashes (1-per-few-days class) and for "when did this start / get fixed" investigations.

  7. AE_ field naming*: every JSON path in the producer's status document becomes an AE_-prefixed journal field (per log2journal --prefix 'AE_' on the ingestion server). See AE_FIELDS.md for the verified map and enum meanings.

Table of contents

DocPurpose
AE_FIELDS.mdVerified field map (~80 rows) + enum meanings for triage. Indispensable.
transports.mdCloud API + direct agent API call patterns; ssh footnote.
update-cadence.mdAfter-the-fact model, dedup, push timing, disk commits, query implications.
query-discipline.mdThe multi-value filter syntax, structured-filters-first rule, anti-patterns.
finding-crashes.mdRecipe: signal crashes (SIGSEGV / SIGBUS / SIGFPE / SIGABRT) on stable.
finding-fatals.mdRecipe: deliberate fatals (OOM, disk full, asserts).
recipes/INDEX.mdLive catalog of recipes (find-by-function, find-by-version, find-related-to-work).
how-tos/INDEX.mdLive catalog: every analysis question becomes a how-to entry.

Live how-to rule (mandatory)

If asked a concrete question about agent-events that requires non-trivial analysis (multiple file reads, running queries, cross-referencing with producer source) AND the answer is not already documented in the per-domain guides above or in recipes/, the assistant MUST author a new how-to under how-tos/<slug>.md and add a one-line entry to how-tos/INDEX.md BEFORE completing the task.

Scripts (in scripts/)

ScriptPurpose
_lib.shHelpers (agentevents_* prefix). Sources query-netdata-agents/scripts/_lib.sh. Token-safe; ships a no-leak self-test.
get-events.shFetch events of interest. Index-friendly defaults. JSON output to .local/audits/query-agent-events/.
analyze-events.shGroup-by stats over a downloaded dump (signal, version, fatal_function, architecture, etc.).
redact-events.shOpt-in redaction (machine_guid / claim_id / host_id / ephemeral_id -> placeholders). For sharing only.

Path discipline

This skill follows <repo>/.agents/sensitive-data-discipline.md:

  • Repo files: repo-relative (<repo>/src/...).
  • Sibling Netdata-org repos: ${NETDATA_REPOS_DIR}/<repo>/....
  • agent-events host / namespace / machine GUID / node ID: ALWAYS via env keys. Never literal values in any committed file.
  • Producer ingest URL: NEVER quoted literally. Reference only as src/daemon/status-file.c:988.
  • Fetched event payloads land under <repo>/.local/audits/query-agent-events/<timestamp>.json (gitignored). Do NOT paste raw event JSON into committed artifacts.

Required env keys

KeyRole
NETDATA_CLOUD_TOKENCloud REST token (long-lived).
NETDATA_CLOUD_HOSTNAMECloud REST API host.
AGENT_EVENTS_HOSTNAMEDual-duty: ssh host AND direct-HTTP host of the ingestion node. Can be IP or DNS name. NOT the journalctl namespace (hardcoded agent-events); NOT the Cloud room name (also hardcoded agent-events).
AGENT_EVENTS_MACHINE_GUIDAgent machine GUID for direct-agent transport.
AGENT_EVENTS_NODE_IDCloud node UUID for cloud-proxy transport.

All values live in <repo>/.env (gitignored). See <repo>/.agents/ENV.md for setup (where each value comes from, sample formats, common mistakes).

  • query-netdata-cloud -- transport: Cloud REST API.
  • query-netdata-agents -- transport: direct agent REST + bearer auto-mint.
  • This skill consumes both via their _lib.sh helpers.