zhangshenghang opened a new pull request, #13282:
URL: https://github.com/apache/seatunnel/pull/13282

   Yes, I am contributing to Apache SeaTunnel.
   
   ## What documentation issues were found
   
   While auditing the docs on `upstream/dev` against the source code, I found 
the following real issues:
   
   1. **False claim about `job-schedule-strategy` with dynamic slots (EN + ZH, 
hybrid + separated deployment guides).** All four deployment guides stated that 
with `dynamic-slot: true`, a `job-schedule-strategy: WAIT` configuration "will 
become invalid and will be forcibly changed to REJECT". No such override exists 
in the engine: `CoordinatorService` (seatunnel-engine-server) reads the 
strategy only from the engine config (`isWaitStrategy = 
scheduleStrategy.equals(ScheduleStrategy.WAIT)`) and, when resources are 
insufficient, `WAIT` sleeps 3 seconds and retries scheduling while `REJECT` 
fails the job immediately. `dynamicSlot` is only used for slot accounting 
(`DefaultSlotService`, `ResourceRequestHandler`). The ZH variants additionally 
contained a copy-paste-breaking typo in the example wording (`dynamic-slot: 
ture`).
   2. **Wrong checkpoint-interval "default" in the tuning guide (EN + ZH).** 
The tuning guide said checkpoints run "every 10s by default", but the engine 
default for `seatunnel.checkpoint.interval` is `300000` ms 
(`ServerConfigOptions.CHECKPOINT_INTERVAL`). The 10 s value only exists because 
the shipped `config/seatunnel.yaml` template sets `checkpoint.interval: 10000`. 
Both facts are now stated correctly.
   3. **Broken OpenAPI reference in the ZH REST API v2 doc.** 
`docs/zh/engines/zeta/rest-api-v2.md` linked the OpenAPI contract with an 
unresolvable `@site/docs/engines/zeta/openapi-v2.yaml` target, while the EN doc 
uses a relative link to the spec that exists only under 
`docs/en/engines/zeta/`. The ZH page now uses the same relative link, and the 
language-neutral spec file `openapi-v2.yaml` is mirrored into 
`docs/zh/engines/zeta/` so the link resolves in the localized docs tree.
   4. **Missing "Key Source Files" reference sections in the ZH architecture 
docs.** `checkpoint-mechanism.md`, `dag-execution.md` and 
`engine-architecture.md` under `docs/zh/architecture/` never included the 
"关键源码文件" reference subsections that the EN versions have (all referenced source 
files were verified to exist in the repo).
   5. **`null_format` option description contradicting the code in 
`OssJindoFile.md` (EN).** The option table claimed the option is "Only used in 
`text` format", but it is applied by both `TextReadStrategy` and 
`CsvReadStrategy` (connector-file-base), i.e. `text` **or** `csv` — which is 
also what the option detail section on the same page and the other file 
connector docs say.
   
   ## What was changed
   
   - Corrected the scheduling-strategy paragraph in the four deployment guides 
and removed the `ture` typo as part of the rewrite.
   - Corrected the checkpoint-interval default in the tuning guides.
   - Fixed the ZH OpenAPI link and added the mirrored `openapi-v2.yaml`.
   - Added the missing ZH "Key Source Files" reference subsections with the 
same file links as EN.
   - Corrected the `null_format` table row to `text` or `csv`.
   
   ## Which areas/files were updated
   
   - `docs/en/engines/zeta/` and `docs/zh/engines/zeta/`: 
`hybrid-cluster-deployment.md`, `separated-cluster-deployment.md`, 
`tuning-guide.md`, `rest-api-v2.md` (+ new `openapi-v2.yaml` under `zh`)
   - `docs/zh/architecture/`: `fault-tolerance/checkpoint-mechanism.md`, 
`engine/dag-execution.md`, `engine/engine-architecture.md`
   - `docs/en/connectors/source/OssJindoFile.md`
   
   In total 12 files (11 modified + 1 new spec copy).
   
   ## English and Chinese docs
   
   Yes — every fix above was applied to both the English and Chinese docs where 
both versions exist, except the mirrored OpenAPI spec (language-neutral file) 
and the `null_format` table row, whose ZH counterpart was already correct.
   
   ## Duplicate PR check (last 7 days)
   
   I searched all PRs opened in the last 7 days (`gh pr list --author 
zhangshenghang`, plus all open docs PRs) and checked their changed-file lists: 
#13274 (Doris/StarRocks/TDengine/Milvus/transforms), #13270 (TextChunk), #13258 
(Couchbase/ES/RocketMQ/Paimon/Rabbitmq/Redis), #13234 
(Clickhouse/InfluxDB/SaaS/CDC docs), #13213 (E2E), and the older open docs PRs 
#13193/#13149/#13097/#13072/#12993/#12986/#12971. None of them touch the 
deployment guides, tuning guide, ZH architecture docs, REST v2 ZH doc, or the 
OssJindoFile `null_format` row. One partial overlap is disclosed: open PR 
#13149 fixes the same class of `null_format` wording mistake in `CosFile.md` 
(EN+ZH) and rewrites other parts of `OssJindoFile.md` (the `schema` option row 
and the `file_format_type` section), but it does not touch the `null_format` 
row fixed here, so this PR complements it without duplicating it.
   
   ## How the change was verified
   
   - A repo-wide internal-link and image-reference scan over all 1114 markdown 
files (before and after the change): the `@site/...` target was the only 
genuine broken link; after the fix, all internal links resolve (the two 
remaining scanner hits are regex literals inside inline code, not links).
   - All source-file links added to the ZH architecture docs were checked to 
exist in the repository.
   - The copied `openapi-v2.yaml` was validated as parseable YAML and is 
byte-identical to the EN version.
   - Verified every corrected claim against the code: `CoordinatorService.java` 
(WAIT/REJECT behavior), `ServerConfigOptions.java` (`checkpoint.interval` 
default 300000, basic-auth defaults), `DefaultSlotService.java` / 
`ResourceRequestHandler.java` (dynamic-slot usage), and `TextReadStrategy.java` 
/ `CsvReadStrategy.java` (`null_format` applies to text and csv).
   - A full documentation site build was not run because the Docusaurus site 
lives in a separate repository; the checks above were run instead.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to