| 1 | # Pipeline |
| 2 | |
| 3 | This document maps the integrations pipeline end to end -- every |
| 4 | script, every input, every output, every CI workflow. All path |
| 5 | citations are repo-relative; line citations refer to the file at |
| 6 | HEAD of `master` at the time this skill was last updated. |
| 7 | |
| 8 | ## The integration documentation pipeline |
| 9 | |
| 10 | ``` |
| 11 | [ YAML sources ] |
| 12 | | |
| 13 | v |
| 14 | +------------------------+ |
| 15 | | gen_integrations.py | (orchestrator, validator, renderer) |
| 16 | +------------------------+ |
| 17 | | |
| 18 | v reads |
| 19 | +------------------------+ |
| 20 | | integrations.js (gitignored) |
| 21 | | integrations.json (gitignored) |
| 22 | +------------------------+ |
| 23 | |
| 24 | [ collector metadata.yaml + taxonomy.yaml + taxonomy registries ] |
| 25 | | |
| 26 | v |
| 27 | +------------------------+ |
| 28 | | gen_taxonomy.py | (taxonomy validator, resolver) |
| 29 | +------------------------+ |
| 30 | | |
| 31 | v |
| 32 | +------------------------+ |
| 33 | | taxonomy.json (gitignored) |
| 34 | +------------------------+ |
| 35 | | |
| 36 | v |
| 37 | +--------------------------------+ |
| 38 | | gen_docs_integrations.py | (per-integration .md files) |
| 39 | +--------------------------------+ |
| 40 | | |
| 41 | v |
| 42 | +--------------------------------+ |
| 43 | | gen_doc_collector_page.py | (src/collectors/COLLECTORS.md) |
| 44 | +--------------------------------+ |
| 45 | | |
| 46 | v |
| 47 | +--------------------------------+ |
| 48 | | gen_doc_secrets_page.py | (src/collectors/SECRETS.md) |
| 49 | +--------------------------------+ |
| 50 | | |
| 51 | v |
| 52 | +--------------------------------+ |
| 53 | | gen_doc_service_discovery_page.py | (src/collectors/SERVICE-DISCOVERY.md) |
| 54 | +--------------------------------+ NOT in CI today -- see gotchas.md |
| 55 | ``` |
| 56 | |
| 57 | All four downstream scripts read the **same** `integrations.js` |
| 58 | (or its data inside; details below). They run sequentially but |
| 59 | do not cross-talk. |
| 60 | |
| 61 | ## Stage 1 -- `gen_integrations.py` (orchestrator) |
| 62 | |
| 63 | Repo path: `integrations/gen_integrations.py`. |
| 64 | |
| 65 | ### Inputs |
| 66 | |
| 67 | - **Categories**: `integrations/categories.yaml` (validated |
| 68 | against `integrations/schemas/categories.json` at |
| 69 | `gen_integrations.py:344-355`). |
| 70 | - **Distros**: `.github/data/distros.yml` (loaded via |
| 71 | `load_yaml` at `gen_integrations.py:1330` -- WITHOUT |
| 72 | validation; the `distros.json` schema exists but is not |
| 73 | consulted -- see `gotchas.md`). |
| 74 | - **Per-integration `metadata.yaml`** files matched by |
| 75 | `METADATA_PATTERN = '*/metadata.yaml'` |
| 76 | (`gen_integrations.py:25`) under nine collector source roots: |
| 77 | |
| 78 | | Root | Integration types served | |
| 79 | |---|---| |
| 80 | | `src/collectors` | C plugins (apps, cgroups, diskspace, ebpf, freebsd, idlejitter, macos, proc, slabinfo, statsd, systemd-journal, tc, timex, xenstat, log2journal, charts.d, python.d) | |
| 81 | | `src/collectors/charts.d.plugin` | shell-based charts.d collectors | |
| 82 | | `src/collectors/python.d.plugin` | Python collectors (am2320, etc.) | |
| 83 | | `src/collectors/guides` | tutorial-style content | |
| 84 | | `src/go/plugin/go.d/collector` | the Go collector tree (the bulk) | |
| 85 | | `src/go/plugin/scripts.d/collector` | scripts.d (shell) | |
| 86 | | `src/go/plugin/ibm.d/modules` | ibm.d collectors (db2, mq, etc.) | |
| 87 | | `src/go/plugin/ibm.d/modules/websphere` | websphere/{jmx,mp,pmi}/ subcollectors -- listed separately because they are 1 level deeper | |
| 88 | | `src/crates/netdata-otel` | the OTEL Rust crate's collector metadata | |
| 89 | |
| 90 | - **Exporters**: `src/exporting/*/metadata.yaml` |
| 91 | (`gen_integrations.py:43`). |
| 92 | - **Agent notifications**: |
| 93 | `src/health/notifications/*/metadata.yaml` (`:47`). |
| 94 | - **Cloud notifications**: |
| 95 | `integrations/cloud-notifications/metadata.yaml` (`:51`). |
| 96 | - **Logs**: `integrations/logs/metadata.yaml` (`:55`). |
| 97 | - **Authentication**: |
| 98 | `integrations/cloud-authentication/metadata.yaml` (`:59`). |
| 99 | - **Secretstore**: |
| 100 | `src/go/plugin/agent/secrets/secretstore/backends/*/metadata.yaml` |
| 101 | (`:63`). |
| 102 | - **Service discovery**: |
| 103 | `src/go/plugin/go.d/discovery/sdext/discoverer/*/metadata.yaml` |
| 104 | (`:67`). |
| 105 | - **Deploy**: `integrations/deploy.yaml` (`:39-41`). |
| 106 | |
| 107 | - **Schemas**: `integrations/schemas/*.json` -- loaded on demand |
| 108 | via `Registry(retrieve=retrieve_from_filesystem)` |
| 109 | (`gen_integrations.py:163-169`). Each integration type has its |
| 110 | own `Draft7Validator` instance (`:171-219`). |
| 111 | |
| 112 | - **Templates**: `integrations/templates/**` -- Jinja env at |
| 113 | `gen_integrations.py:230-241`. Custom delimiters: `[[ ]]` |
| 114 | for variables and `[% %]` for control statements (so that |
| 115 | the template can pass through embedded `{% ... %}` and |
| 116 | `{{ ... }}` markers untouched). See `gotchas.md`. |
| 117 | |
| 118 | ### Validation behavior |
| 119 | |
| 120 | For each integration type, `gen_integrations.py` runs a |
| 121 | `Draft7Validator.validate(...)` call (e.g. `:350`, `:372`, |
| 122 | `:399`, `:437`, `:485`, `:533`, `:581`, `:629`, `:677`, `:725`). |
| 123 | On any `ValidationError`, the script calls `warn(...)`. |
| 124 | **Warnings are fatal**: `fail_on_warnings()` (`:150-160`) |
| 125 | returns 1, causing the CI workflow to fail and abort doc |
| 126 | regeneration. |
| 127 | |
| 128 | The validator IS strict about declared properties; it is NOT |
| 129 | strict about extra properties (no `additionalProperties: false` |
| 130 | on collector.json). Unknown keys (`alternative_monitored_instances`, |
| 131 | `most_popular`) pass through silently. They appear in |
| 132 | `integrations.js` but no template renders them. See `gotchas.md`. |
| 133 | |
| 134 | ### Rendering behavior |
| 135 | |
| 136 | For each integration type, the script: |
| 137 | |
| 138 | 1. Loads the YAML(s). |
| 139 | 2. Validates each entry against the type's JSON Schema. |
| 140 | 3. Calls `make_id` (collectors only -- `:766`, |
| 141 | `f'{plugin}-{module}-{instance}'`). |
| 142 | 4. Computes `edit_link` from `_src_path` (`:777`). |
| 143 | 5. Sorts by id/path/index. |
| 144 | 6. Calls `dedupe_integrations` (`:789`); duplicate ids yield |
| 145 | warnings. |
| 146 | 7. Renders every section listed in `*_RENDER_KEYS` (`:71-122`) |
| 147 | through Jinja, storing the result back on the item under |
| 148 | that key. Sections come from the type's schema (e.g. |
| 149 | `COLLECTOR_RENDER_KEYS = ['alerts', 'metrics', 'functions', |
| 150 | 'overview', 'related_resources', 'setup', |
| 151 | 'troubleshooting']` at `:71`). |
| 152 | 8. Each section is rendered TWICE -- with `clean=False` (rich |
| 153 | variant for the JS / cloud-frontend output) and `clean=True` |
| 154 | (clean variant for the JSON / GitHub-rendered `.md` |
| 155 | output). Both variants are kept in parallel `clean_*` |
| 156 | lists. |
| 157 | 9. Strips internal-only keys (`_src_path`, `_repo`, `_index`) |
| 158 | before serialization. |
| 159 | |
| 160 | ### Two-pass templating with `meta.variables` |
| 161 | |
| 162 | When a metadata entry declares |
| 163 | `meta.monitored_instance.variables` (collectors) or |
| 164 | `meta.variables` (other types), the FIRST pass produces |
| 165 | markdown that may still contain `[[ variables.foo ]]` markers. |
| 166 | The renderer detects this with a regex and performs a SECOND |
| 167 | Jinja pass over the rendered string with `variables=...` in |
| 168 | context (`:930-934`). This lets metadata authors inject |
| 169 | runtime-style placeholders into rendered text. |
| 170 | |
| 171 | **Divergence**: collectors look up |
| 172 | `monitored_instance.variables`; exporters and notifications |
| 173 | look up `meta.variables` directly. Same goal, different lookup |
| 174 | path -- a known wart. |
| 175 | |
| 176 | ### Outputs |
| 177 | |
| 178 | Two files are written (`gen_integrations.py:1311-1325`): |
| 179 | |
| 180 | - `integrations/integrations.js` -- assembles the |
| 181 | `integrations/templates/integrations.js` Jinja shell with |
| 182 | `categories=...` and `integrations=...` JSON, then runs |
| 183 | `convert_local_links` to rewrite any `](/...)` in the body to |
| 184 | absolute GitHub URLs at `https://github.com/netdata/netdata/blob/master/...`. |
| 185 | The first 2 lines are a banner: |
| 186 | ``` |
| 187 | // DO NOT EDIT THIS FILE DIRECTLY |
| 188 | // It gets generated by integrations/gen_integrations.py in the Netdata repo |
| 189 | ``` |
| 190 | The body is `export const categories = [...]; export const |
| 191 | integrations = [...]`. |
| 192 | - `integrations/integrations.json` -- pure JSON with the |
| 193 | `clean` variant of `{categories, integrations}`. No banner. |
| 194 | |
| 195 | Both are gitignored (`.gitignore:162-163`). They are produced |
| 196 | fresh on every run; in CI, the workflow `rm`s them after the |
| 197 | downstream scripts read them so they are NOT included in the |
| 198 | auto-PR. |
| 199 | |
| 200 | ## Parallel taxonomy stage -- `gen_taxonomy.py` |
| 201 | |
| 202 | Repo path: `integrations/gen_taxonomy.py`. |
| 203 | |
| 204 | ### Inputs |
| 205 | |
| 206 | - Collector `metadata.yaml` files from the same collector source |
| 207 | roots used by `gen_integrations.py`. |
| 208 | - Sibling collector `taxonomy.yaml` files discovered as |
| 209 | `<collector>/taxonomy.yaml`. |
| 210 | - `integrations/taxonomy/sections.yaml` -- the stable section |
| 211 | registry. Collector files reference only `section_id:`, never |
| 212 | `section_path:`. |
| 213 | - `integrations/taxonomy/icons.yaml` -- allowed icon ids. |
| 214 | - `integrations/schemas/taxonomy_collector.json`, |
| 215 | `taxonomy_sections.json`, and `taxonomy_output.json`. |
| 216 | |
| 217 | ### Validation behavior |
| 218 | |
| 219 | The generator validates closed v1 authoring schemas, checks that |
| 220 | literal owned contexts and widget references resolve to real contexts |
| 221 | in the owning collector's `metadata.yaml`, and requires dynamic |
| 222 | selectors to be declared by the owning collector: |
| 223 | |
| 224 | - `context_prefix:` requires |
| 225 | `metrics.dynamic_context_prefixes: [{prefix, reason}]`; taxonomy may |
| 226 | use a narrower prefix under the declared namespace. |
| 227 | - `collect_plugin:` requires |
| 228 | `metrics.dynamic_collect_plugins: [{plugin, reason}]`. |
| 229 | |
| 230 | Findings render as plain text locally and as GitHub Actions |
| 231 | annotations in CI. Fatal findings fail the run. |
| 232 | |
| 233 | ### Outputs |
| 234 | |
| 235 | `integrations/taxonomy.json` is written by default and validated |
| 236 | against `integrations/schemas/taxonomy_output.json`. The file is |
| 237 | gitignored and removed by `generate-integrations.yml` cleanup, just |
| 238 | like `integrations/integrations.js` and `integrations/integrations.json`. |
| 239 | |
| 240 | Run validation only: |
| 241 | |
| 242 | ```bash |
| 243 | python3 integrations/gen_taxonomy.py --check-only |
| 244 | ``` |
| 245 | |
| 246 | Seed a collector taxonomy from existing metadata contexts: |
| 247 | |
| 248 | ```bash |
| 249 | python3 integrations/gen_taxonomy_seed.py src/go/plugin/go.d/collector/apache/metadata.yaml --module-name apache --section-id applications.apache --placement-id apache --icon apache |
| 250 | ``` |
| 251 | |
| 252 | The seed helper emits a flat `items:` tree. For collectors with richer |
| 253 | dashboard layout needs, convert that flat list into explicit |
| 254 | `owned_context`, `group`, `flatten`, `selector`, `context`, `grid`, |
| 255 | `first_available`, or `view_switch` items before opening the PR. |
| 256 | |
| 257 | Pull-request coverage is checked by: |
| 258 | |
| 259 | ```bash |
| 260 | python3 integrations/check_collector_taxonomy.py --pr-diff origin/master...HEAD |
| 261 | ``` |
| 262 | |
| 263 | ### Commands a maintainer runs locally |
| 264 | |
| 265 | ```bash |
| 266 | cd <repo> |
| 267 | ./integrations/pip.sh # installs jsonschema referencing jinja2 ruamel.yaml |
| 268 | python3 integrations/gen_integrations.py |
| 269 | ``` |
| 270 | |
| 271 | Run from the repo root. The script depends on relative paths |
| 272 | hard-coded in `gen_integrations.py:11-37`. |
| 273 | |
| 274 | ## Stage 2 -- `gen_docs_integrations.py` |
| 275 | |
| 276 | Repo path: `integrations/gen_docs_integrations.py`. |
| 277 | |
| 278 | ### Inputs |
| 279 | |
| 280 | - `integrations/integrations.js` -- the script parses it by |
| 281 | string-splitting on `export const categories = ` and |
| 282 | `export const integrations = ` (`:129-140`). It does NOT |
| 283 | read `integrations.json`. |
| 284 | |
| 285 | ### Outputs |
| 286 | |
| 287 | For each integration entry, the script writes either a |
| 288 | `<plugin-dir>/integrations/<slug>.md` file or a |
| 289 | `<plugin-dir>/README.md` file (depending on type). The full |
| 290 | mapping per integration type is in `per-type-matrix.md`. Slug |
| 291 | rules are in `artifacts-and-banners.md`. |
| 292 | |
| 293 | After writing, the script: |
| 294 | |
| 295 | 1. Calls `resolve_related_links()` (`:56-78`) to convert |
| 296 | `{% relatedResource id="..." %}name{% /relatedResource %}` |
| 297 | markers (left in by `templates/overview/collector.md:42` |
| 298 | and `templates/related_resources.md:5`) into |
| 299 | `[name](/path)` markdown links. **Two-pass resolution**: |
| 300 | the markers are present in pass 1; they get rewritten in |
| 301 | pass 2 after every file is written so the id-to-path map |
| 302 | is complete. If the id is not found, the marker is |
| 303 | replaced with bare `name` text (silent fallback). |
| 304 | |
| 305 | 2. Calls `make_symlinks(symlink_dict)` (`:527-544`) to symlink |
| 306 | `<plugin-dir>/README.md -> integrations/<sole-file>.md` |
| 307 | when the directory holds exactly one integration. Only |
| 308 | fires when `len(list(integrations_dir.iterdir())) == 1` |
| 309 | (`:466`). Multi-integration directories are NOT |
| 310 | symlinked. |
| 311 | |
| 312 | 3. Cleans the corresponding `**/integrations` directories |
| 313 | BEFORE writing (`:19-41`), so removed integrations vanish |
| 314 | from the tree. |
| 315 | |
| 316 | ### Scoped regen |
| 317 | |
| 318 | The script accepts `-c plugin/module` to scope cleanup and |
| 319 | regen to one collector (`:578-583`). Useful locally: |
| 320 | |
| 321 | ```bash |
| 322 | python3 integrations/gen_docs_integrations.py -c go.d.plugin/snmp |
| 323 | ``` |
| 324 | |
| 325 | NOT used by CI; CI always runs without `-c` (full regen). |
| 326 | |
| 327 | ## Stage 3 -- `gen_doc_collector_page.py` |
| 328 | |
| 329 | Repo path: `integrations/gen_doc_collector_page.py`. |
| 330 | |
| 331 | Reads `integrations/integrations.js` (`:38-47`). Walks the |
| 332 | category tree; the "section-level" categories are normally |
| 333 | children of `data-collection`, plus the top-level `flows` |
| 334 | category (`:82-86`). `flows` is deliberately included because |
| 335 | Network Flows entries cover both flow protocols and enrichment |
| 336 | inputs, so the Monitor Anything page must list them together |
| 337 | under a `Network Flows` section instead of dropping them into |
| 338 | `Other`. |
| 339 | |
| 340 | Writes `src/collectors/COLLECTORS.md` (committed). This is the |
| 341 | "Monitor anything with Netdata" umbrella marketing page that |
| 342 | lists every collector and Network Flows integration in tabular |
| 343 | form, grouped by section. The write path is |
| 344 | `generate_collectors_md()` (`:565-584`), which renders the |
| 345 | header plus dynamic tables and atomically replaces the file. |
| 346 | |
| 347 | ### Notable behaviors |
| 348 | |
| 349 | - Sort order: "Linux first, Other last" (`:285-301`); Network |
| 350 | Flows follows its position in `integrations/categories.yaml` |
| 351 | because it is treated as a section. |
| 352 | - Description extraction: `extract_description_from_overview` |
| 353 | reads `## Overview` body, uses the first sentence (`:143-183`); |
| 354 | falls back to `meta.monitored_instance.description`; final |
| 355 | fallback `Monitor <name>`. Because this text becomes the |
| 356 | Monitor Anything table description, the first sentence of the |
| 357 | overview must describe the integration itself, not a setting, |
| 358 | variable, default, limit, or troubleshooting detail. See |
| 359 | `description-authoring.md`. |
| 360 | - Slug for table links: `to_slug(display_name)` -- lowercase, |
| 361 | spaces to `_`, `/` to `-`, strips parentheses (`:213-215`). |
| 362 | - Hardcoded marketing anchors: `_render_tech_navigation` |
| 363 | (`:424-493`) writes `#cloud-provider-managed`, `#kubernetes`, |
| 364 | `#search-engines`, `#freebsd`, `#message-brokers`, etc. |
| 365 | Several of these category IDs do NOT exist in |
| 366 | `categories.yaml` -- some links go to non-existent anchors. |
| 367 | See `gotchas.md`. Header literal "850+ integrations" is also |
| 368 | baked in. |
| 369 | |
| 370 | ## Stage 4 -- `gen_doc_secrets_page.py` |
| 371 | |
| 372 | Repo path: `integrations/gen_doc_secrets_page.py`. |
| 373 | |
| 374 | Reads `integrations/integrations.js` (`:213-217`), filters |
| 375 | entries where `integration_type == 'secretstore'`, builds the |
| 376 | "Supported Secretstore Backends" table from each backend's |
| 377 | `meta.kind`, `meta.name`, `collector_configs.summary.{operand_format, |
| 378 | example_operand}`, and renders via |
| 379 | `integrations/templates/secrets.md`. Writes |
| 380 | `src/collectors/SECRETS.md` (committed, `:358`). |
| 381 | |
| 382 | The bulk of `SECRETS.md` is **static content baked into the |
| 383 | script** (`SECRETS_PAGE` dict, `:20-203`). Only the backends |
| 384 | table is dynamic. To change the static prose, edit the script. |
| 385 | |
| 386 | ## Stage 5 -- `gen_doc_service_discovery_page.py` |
| 387 | |
| 388 | Repo path: `integrations/gen_doc_service_discovery_page.py`. |
| 389 | |
| 390 | Mirror of the secrets stage for service discovery. Reads |
| 391 | `integrations.js`, filters |
| 392 | `integration_type == 'service_discovery'`, renders via |
| 393 | `integrations/templates/service_discovery.md`. Writes |
| 394 | `src/collectors/SERVICE-DISCOVERY.md` (committed, `:382`). |
| 395 | Most content is static (`SD_PAGE` dict, `:21-257`). |
| 396 | |
| 397 | **KNOWN GAP**: this stage is NOT wired into the |
| 398 | `generate-integrations.yml` workflow. CI does not run it. The |
| 399 | file in tree drifts from metadata.yaml until a developer runs |
| 400 | the script manually (or a future PR adds it to CI). See |
| 401 | `gotchas.md` and the SOW followups. |
| 402 | |
| 403 | ## CI workflow 1 -- `generate-integrations.yml` |
| 404 | |
| 405 | Repo path: `.github/workflows/generate-integrations.yml`. |
| 406 | |
| 407 | ### Triggers |
| 408 | |
| 409 | - `push` to `master` filtered by paths |
| 410 | (`generate-integrations.yml:6-25`): |
| 411 | - `**/metadata.yaml` (every collector / exporter / notification |
| 412 | metadata) |
| 413 | - `integrations/templates/**` |
| 414 | - `integrations/schemas/**` |
| 415 | - `integrations/categories.yaml`, `integrations/deploy.yaml` |
| 416 | - `integrations/cloud-notifications/metadata.yaml`, |
| 417 | `integrations/cloud-authentication/metadata.yaml` |
| 418 | - the four older Python scripts (NOT |
| 419 | `gen_doc_service_discovery_page.py` -- the gap) |
| 420 | - `workflow_dispatch` -- manual. |
| 421 | |
| 422 | ### Concurrency |
| 423 | |
| 424 | - `integrations-${{ github.ref }}`, `cancel-in-progress: true`. |
| 425 | |
| 426 | ### Repo gate |
| 427 | |
| 428 | - `if: github.repository == 'netdata/netdata'` -- forks do NOT |
| 429 | trigger this workflow. |
| 430 | |
| 431 | ### Steps |
| 432 | |
| 433 | 1. `actions/checkout@v6` (depth 1, recursive submodules). |
| 434 | 2. `apt install python3-venv` + `./integrations/pip.sh` to |
| 435 | install Python deps. |
| 436 | 3. `python3 integrations/gen_integrations.py`. |
| 437 | 4. `python3 integrations/gen_docs_integrations.py`. |
| 438 | 5. `python3 integrations/gen_doc_collector_page.py`. |
| 439 | 6. `python3 integrations/gen_doc_secrets_page.py`. |
| 440 | 7. **NOT** `gen_doc_service_discovery_page.py` -- gap. |
| 441 | 8. `rm -rf go.d.plugin virtualenv integrations/integrations.js |
| 442 | integrations/integrations.json` -- prevents the auto-PR from |
| 443 | committing the runtime artifacts. |
| 444 | 9. `peter-evans/create-pull-request@v8` -- branch |
| 445 | `integrations-regen`, label `integrations-update`, title |
| 446 | `Regenerate integrations docs`, token |
| 447 | `NETDATABOT_GITHUB_TOKEN`. Reviewed and merged manually. |
| 448 | 10. Slack failure notification on master failures. |
| 449 | |
| 450 | ## CI workflow 2 -- `check-markdown.yml` |
| 451 | |
| 452 | Repo path: `.github/workflows/check-markdown.yml`. |
| 453 | |
| 454 | ### Triggers |
| 455 | |
| 456 | - `pull_request` filtered by paths: |
| 457 | - `**/*.md`, `**/*.mdx` |
| 458 | - `docs/**`, `**/metadata.yaml`, `integrations/**` |
| 459 | |
| 460 | ### Steps |
| 461 | |
| 462 | 1. Checkout PR branch and the `netdata/learn` repo. |
| 463 | 2. Install Python deps (`./integrations/pip.sh`). |
| 464 | 3. Run `gen_integrations.py`, `gen_docs_integrations.py`, |
| 465 | `gen_doc_collector_page.py`, `gen_doc_secrets_page.py` |
| 466 | (same gap on SD page generator). |
| 467 | 4. Run `learn/ingest/ingest.py --local-repo netdata:... |
| 468 | --ignore-on-prem-repo --fail-links-netdata` |
| 469 | (`check-markdown.yml:64-69`) -- validates that all |
| 470 | generated markdown links resolve through Learn's ingest |
| 471 | pipeline. |
| 472 | |
| 473 | This workflow validates but does NOT auto-commit. It acts as |
| 474 | a gate on PRs. A failure here means a PR cannot merge until |
| 475 | the metadata or links are fixed. |
| 476 | |
| 477 | ## CMake target -- `render-docs` |
| 478 | |
| 479 | Repo path: `packaging/cmake/Modules/NetdataRenderDocs.cmake`. |
| 480 | |
| 481 | A developer-facing convenience target. When wired up by the |
| 482 | build system, it runs the same generator chain (with |
| 483 | `gen_integrations` + `gen_docs_integrations` only by default). |
| 484 | Useful for local validation. NOT a substitute for running the |
| 485 | scripts directly during active development. |
| 486 | |
| 487 | ## End-to-end: a single PR's flow |
| 488 | |
| 489 | 1. Developer edits `src/go/plugin/go.d/collector/foo/metadata.yaml` |
| 490 | (and any other affected consistency-rule files: |
| 491 | `taxonomy.yaml`, `config_schema.json`, stock conf, |
| 492 | `health.d/foo.conf`, `README.md`). |
| 493 | 2. Developer runs locally: |
| 494 | ```bash |
| 495 | ./integrations/pip.sh |
| 496 | python3 integrations/gen_integrations.py |
| 497 | python3 integrations/gen_taxonomy.py --check-only |
| 498 | python3 integrations/check_collector_taxonomy.py --pr-diff master...HEAD |
| 499 | python3 -m unittest integrations.tests.test_taxonomy |
| 500 | python3 integrations/gen_docs_integrations.py -c go.d.plugin/foo |
| 501 | python3 integrations/gen_doc_collector_page.py |
| 502 | python3 integrations/gen_doc_secrets_page.py |
| 503 | ``` |
| 504 | 3. Developer commits the regenerated `integrations/foo.md`, |
| 505 | the symlinked `README.md` (if applicable), and the updated |
| 506 | `src/collectors/COLLECTORS.md` if the collector list |
| 507 | changed. |
| 508 | 4. PR is opened. `check-markdown.yml` runs, regenerates the |
| 509 | same files in CI, and validates Learn ingest. If the dev's |
| 510 | committed files differ from CI's regen, the PR fails. |
| 511 | 5. Reviewer checks collector consistency, including taxonomy |
| 512 | coverage for changed chart contexts. |
| 513 | 6. PR merges. `generate-integrations.yml` triggers on master, |
| 514 | regenerates everything, and opens an `integrations-regen` |
| 515 | PR if anything is now stale (typically nothing, because the |
| 516 | dev already committed the regen). Maintainer merges. |
| 517 | 7. Cloud-frontend's own CI (in |
| 518 | `${NETDATA_REPOS_DIR}/dashboard/cloud-frontend/`) runs |
| 519 | `gen_integrations.py` against the new master and copies |
| 520 | `integrations.js` into its source. See `in-app-contract.md`. |
| 521 | 8. Learn's ingest pulls the new `integrations/foo.md` on its |
| 522 | 3-hourly schedule. See the `learn-site-structure` skill. |
| 523 | |
| 524 | ## End-to-end: Monitor Anything / `COLLECTORS.md` |
| 525 | |
| 526 | 1. `metadata.yaml` entries declare |
| 527 | `meta.monitored_instance.categories`. |
| 528 | 2. `python3 integrations/gen_integrations.py` validates those |
| 529 | categories against `integrations/categories.yaml`, renders the |
| 530 | integration content, and writes the runtime |
| 531 | `integrations/integrations.js` catalog. |
| 532 | 3. `python3 integrations/gen_doc_collector_page.py` reads |
| 533 | `integrations/integrations.js`, groups integrations by Monitor |
| 534 | Anything section, and writes `src/collectors/COLLECTORS.md`. |
| 535 | 4. `check-markdown.yml` runs the same generator before Learn |
| 536 | ingest on PRs, so broken generated `COLLECTORS.md` content |
| 537 | (for example, unresolved links) blocks the PR. It does not |
| 538 | diff-check that the committed `COLLECTORS.md` file is fresh. |
| 539 | 5. `generate-integrations.yml` runs the same generator after |
| 540 | metadata changes land on `master` and opens the |
| 541 | `integrations-regen` PR if committed generated artifacts drift. |
| 542 | 6. Learn ingests `src/collectors/COLLECTORS.md` as the |
| 543 | "Monitor anything with Netdata" page. |
| 544 | |
| 545 | For Network Flows specifically, keep the top-level `flows` |
| 546 | category handling in `gen_doc_collector_page.py`. Without it, |
| 547 | NetFlow / IPFIX / sFlow and enrichment entries will not appear |
| 548 | as a coherent `Network Flows` section on Monitor Anything. |