morningman opened a new pull request, #4060:
URL: https://github.com/apache/doris-website/pull/4060

   ## Purpose
   
   Audit the [4.0.8 release 
notes](https://github.com/apache/doris/issues/66642) against the 4.x Chinese 
docs, so the docs are calibrated when 4.0.8 ships.
   
   Every item was verified against the actual PR diff on `branch-4.0` rather 
than the release-note wording alone. That caught one inaccuracy: the release 
note says the BE `_stream_load_forward` endpoint "is now registered only when 
the configuration is enabled", but the code always registers it and rejects 
requests with 403 at runtime. The docs describe the actual behavior.
   
   Only the 4.x **Chinese** docs are touched. The single English file added is 
the counterpart of the new API doc — the shared sidebar resolves doc IDs 
against the default locale, so the entry would break the build without it.
   
   ## Behavior changes (all annotated with 4.0.8)
   
   | PR | Doc | Change |
   |---|---|---|
   | #66604 | `admin-manual/cluster-management/upgrade.md` | New section on the 
`variable_version=400` migration; `enable_nereids_distribute_planner` is now 
enabled on upgrade from 3.x, with a rollback note |
   | #61288 | `config/be-config.md` | New 
`enable_compaction_pause_on_high_memory`, default `true` to `false` |
   | #64935 | `be-config.md`, `fe-config.md`, `group-commit-manual.md` | New 
BE/FE `enable_group_commit_streamload_be_forward`; `_stream_load_forward` 
returns 403 when disabled and requires global `LOAD` when enabled |
   | #63729 | `fe-config.md`, `data-bucketing.md` | New 
`autobucket_min_buckets` (1 to 3) and `autobucket_max_buckets`; bounds table in 
the auto-bucketing section |
   | #66150 | `compute-storage-decoupled/recycler.md` | Two `*_before_recycle` 
configs, default `true` to `false` |
   | #60887 | `information_schema/partitions.md`, `backend_tablets.md` | Cloud 
mode reports data size as remote size; `LOCAL_DATA_SIZE` and 
`REPLICA_ALLOCATION` column notes |
   | #60944 | `SHOW-PROC.md` | `ReplicaAllocation` is `null` in cloud mode; 
`tablet_health` counts everything as `HEALTHY` in cloud mode |
   
   ## New features
   
   - **New doc** `admin-manual/open-api/fe-http/meta-backup-action.md` (#60739) 
— the cloud metadata backup APIs (`sync_cloud_meta` / `export_meta`), with 
prerequisites, per-API parameter and response tables, the export directory 
layout, the full backup procedure, and a troubleshooting table. Added to 
`versioned_sidebars/version-4.x-sidebars.json`.
   - `file-cache.md` — version note for 
`enable_file_cache_write_index_file_only` (#64995); new section for 
`enable_topn_lazy_mat_phase2_no_write_file_cache` (#65021)
   - `fe-config.md` — `fe_meta_auth_token` (#65551)
   - `node-action.md` — the node add/drop/decommission APIs now require global 
`ADMIN` (#65042, #59708)
   - `SHOW-ROUTINE-LOAD.md` — Kafka sensitive property masking, with the 
matching rules (#64786)
   - `metrics.md` — `doris_fe_virtual_compute_group_switch_total` (#63036)
   - `read-write-separation.md` — warm-up timestamps changed from `HH:mm:ss` to 
`yyyy-MM-dd HH:mm:ss` (#64606)
   - `separating-storage-compute-deploy-manually.md` — `CONF_FILES` multi 
config file support (#63924)
   
   ## Existing content corrected against the current code
   
   The release notes touch these areas, so the surrounding content was checked 
too. Several things were already stale:
   
   - `SHOW-ROUTINE-LOAD.md` — `IsMultiTbl` to `IsMultiTable`; four columns were 
missing (`User`, `Comment`, `ComputeGroup`, `FirstErrorMsg`)
   - `SHOW-PROC.md` — `/colocation_group` documented `ReplicationNum`, the 
column is `ReplicaAllocation`; `ErrorMsg` was missing; the 
`/colocation_group/{id}` columns vary by deployment mode
   - `routine_load_job.md` — added `COMPUTE_GROUP` and `FIRST_ERROR_MSG`, 
documented the master FE routing (#65942)
   - `from-base64.md` — the doc claimed whitespace is ignored; it actually 
yields NULL. Added the "length must be a multiple of 4" rule (#65141)
   - `data-faq.md`, `check-tablet-segment.md` — the `failed to initialize 
storage reader` prefix is gone in 4.0.8 (#64755)
   - `stream-load-manual.md` — `compute_group` header precedence, planning 
bound to the receiving BE's compute group, and the new privilege check (#65571)
   - `iceberg-catalog.mdx`, `paimon-catalog.mdx` — mixed-case column names now 
match case-insensitively (#65094)
   
   ## Not covered
   
   A few release-note items have no home in the current 4.x docs. Listing them 
rather than inventing a place:
   
   - `SHOW TABLETS FROM <table>` has no dedicated SQL doc, so the #60887 
size-semantics change is documented in `partitions.md` only.
   - #65508 (S3 SlowDown retry by role) and #65625 (get-version timeout) are 
Meta Service / Recycler configs. There is no MS config reference doc in 4.x, 
and `max_s3_client_retry` / `s3_client_retry_slow_down` were never documented.
   - #66371 removes `plan_nereids_dump`, which no doc references — nothing to 
delete. `enable_nereids_distribute_planner` was likewise never documented; a 
session-variable reference doc would be a separate change.
   
   Purely internal performance and stability fixes (#64933, #64647, #64638, 
#65174, #64884, and others) need no doc change.
   
   🤖 Generated with [Claude Code](https://claude.com/claude-code)
   
   https://claude.ai/code/session_012KbhQDEMeUvDbwECz6K3bZ
   


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to