This is an automated email from the ASF dual-hosted git repository.
davidzollo pushed a commit to branch dev
in repository https://gitbox.apache.org/repos/asf/seatunnel.git
The following commit(s) were added to refs/heads/dev by this push:
new c685cdeff4 [Docs][Connector-V2] Improve Druid GraphQL and Hive docs
(#11327)
c685cdeff4 is described below
commit c685cdeff439ccd4983288f2fba679b351b53698
Author: Daniel Carter <[email protected]>
AuthorDate: Sat Jul 4 21:59:42 2026 +0800
[Docs][Connector-V2] Improve Druid GraphQL and Hive docs (#11327)
Co-authored-by: DanielCarter-stack
<[email protected]>
---
docs/en/connectors/sink/Druid.md | 4 +++
docs/en/connectors/sink/GraphQL.md | 59 ++++++++++++++++++++++++++++++++++++
docs/en/connectors/sink/Hive.md | 6 +++-
docs/en/connectors/source/GraphQL.md | 2 ++
docs/zh/connectors/sink/Druid.md | 4 +++
docs/zh/connectors/sink/GraphQL.md | 59 ++++++++++++++++++++++++++++++++++++
docs/zh/connectors/sink/Hive.md | 4 +++
docs/zh/connectors/source/GraphQL.md | 2 ++
8 files changed, 139 insertions(+), 1 deletion(-)
diff --git a/docs/en/connectors/sink/Druid.md b/docs/en/connectors/sink/Druid.md
index d6c59049db..2fbbca640d 100644
--- a/docs/en/connectors/sink/Druid.md
+++ b/docs/en/connectors/sink/Druid.md
@@ -57,6 +57,8 @@ The number of rows buffered before SeaTunnel sends one
indexing task to Druid. T
SeaTunnel also flushes the remaining buffered rows when the writer closes.
+Use a larger value for high-throughput batch jobs to reduce the number of
Druid indexing tasks. Use a smaller value when you need data to be submitted to
Druid earlier, but note that this can create more indexing tasks.
+
### common options
Sink plugin common parameters, please refer to [Sink Common
Options](../common-options/sink-common-options.md) for details.
@@ -71,6 +73,8 @@ The connector adds a processing-time column named `timestamp`
to satisfy Druid's
Rows are flushed when the buffered row count reaches `batchSize`. Remaining
rows are also flushed when the writer closes. There is no time-based periodic
flush option.
+The sink is designed for append-style batch ingestion. It does not interpret
CDC update/delete row kinds as Druid upserts or deletes.
+
## Example
### Write One Table
diff --git a/docs/en/connectors/sink/GraphQL.md
b/docs/en/connectors/sink/GraphQL.md
index 0852316125..b0a5329b53 100644
--- a/docs/en/connectors/sink/GraphQL.md
+++ b/docs/en/connectors/sink/GraphQL.md
@@ -56,6 +56,7 @@ It can be downloaded via install-plugin.sh or from Maven
central repository.
- Field names in the upstream row should match variable names used by the
mutation.
- If `variables` already contains a key and `valueCover = true`, the
configured variable value is kept.
- If `valueCover = false`, row field values replace variables with the same
name.
+- The sink sends one mutation request for each input row. If the upstream
source contains multiple tables, make sure the same mutation and variable names
can handle every table routed to this sink.
## Task Example
@@ -155,6 +156,64 @@ sink {
}
```
+### Write Multiple Upstream Tables
+
+```hocon
+source {
+ FakeSource {
+ plugin_output = "fake"
+ tables_configs = [
+ {
+ schema = {
+ table = "graphql_sink_1"
+ fields {
+ id = int
+ val_bool = boolean
+ val_string = string
+ }
+ }
+ rows = [
+ {
+ kind = INSERT
+ fields = [1, true, "NEW"]
+ }
+ ]
+ },
+ {
+ schema = {
+ table = "graphql_sink_2"
+ fields {
+ id = int
+ val_bool = boolean
+ val_string = string
+ }
+ }
+ rows = [
+ {
+ kind = INSERT
+ fields = [2, true, "READY"]
+ }
+ ]
+ }
+ ]
+ }
+}
+
+sink {
+ GraphQL {
+ plugin_input = "fake"
+ url = "http://graphql:8080/v1/graphql";
+ query = """
+ mutation MyMutation($id: Int!, $val_bool: Boolean!, $val_string:
String!) {
+ insert_sink(objects: {id: $id, val_bool: $val_bool, val_string:
$val_string}) {
+ affected_rows
+ }
+ }
+ """
+ }
+}
+```
+
## Changelog
<ChangeLog />
diff --git a/docs/en/connectors/sink/Hive.md b/docs/en/connectors/sink/Hive.md
index 5fd55a3a10..0fdcb5c230 100644
--- a/docs/en/connectors/sink/Hive.md
+++ b/docs/en/connectors/sink/Hive.md
@@ -46,7 +46,7 @@ By default, we use 2PC commit to ensure `exactly-once`
| krb5_path | string | no | /etc/krb5.conf |
| kerberos_principal | string | no | - |
| kerberos_keytab_path | string | no | - |
-| abort_drop_partition_metadata | boolean | no | true |
+| abort_drop_partition_metadata | boolean | no | false |
| parquet_avro_write_timestamp_as_int96 | boolean | no | false |
| overwrite | boolean | no | false |
| data_save_mode | enum | no | APPEND_DATA |
@@ -100,6 +100,8 @@ The keytab path of kerberos
Flag to decide whether to drop partition metadata from Hive Metastore during
an abort operation. Note: this only affects the metadata in the metastore, the
data in the partition will always be deleted(data generated during the
synchronization process).
+The default value is `false`.
+
### parquet_avro_write_timestamp_as_int96 [boolean]
Support writing Parquet INT96 from a timestamp, only valid for parquet files.
@@ -121,6 +123,8 @@ Select how to handle existing data on the target before
writing new data.
Note: overwrite=true and data_save_mode=DROP_DATA are equivalent. Use either
one; do not set both.
+For batch jobs, use either `overwrite = true` or `data_save_mode =
"DROP_DATA"` when the target Hive table should be replaced by the current run.
For normal append jobs, keep the default `data_save_mode = "APPEND_DATA"`.
+
### schema_save_mode [enum]
Before starting the synchronization task, different processing schemes are
selected for the existing table structure on the target side.
diff --git a/docs/en/connectors/source/GraphQL.md
b/docs/en/connectors/source/GraphQL.md
index b8df97da35..4f5f5b61b7 100644
--- a/docs/en/connectors/source/GraphQL.md
+++ b/docs/en/connectors/source/GraphQL.md
@@ -65,6 +65,8 @@ It can be downloaded via install-plugin.sh or from Maven
central repository.
- In subscription mode, set `enable_subscription = true` and use a `ws://` or
`wss://` URL.
- Source queries cannot be GraphQL `mutation` operations.
- `content_field` is usually needed because GraphQL responses are commonly
wrapped under `data`.
+- For streaming query mode over HTTP, configure `poll_interval_millis` to
control how often SeaTunnel sends the same query again.
+- For WebSocket subscription mode, `max_retries` and `retry_delay_ms` only
control reconnect behavior after a subscription connection fails.
## Task Example
diff --git a/docs/zh/connectors/sink/Druid.md b/docs/zh/connectors/sink/Druid.md
index 08f9ec45a6..275cc5878f 100644
--- a/docs/zh/connectors/sink/Druid.md
+++ b/docs/zh/connectors/sink/Druid.md
@@ -57,6 +57,8 @@ SeaTunnel 缓存多少行之后向 Druid 提交一次索引任务。默认值为
写入器关闭时，SeaTunnel 也会把剩余缓存数据提交到 Druid。
+批量写入量较大时，可以适当调大该值，减少提交给 Druid 的索引任务数量。如果希望数据更早提交到 Druid，可以适当调小该值，但这会产生更多索引任务。
+
### common options
Sink 插件通用参数，详见 [Sink Common Options](../common-options/sink-common-options.md)。
@@ -71,6 +73,8 @@ Druid 写入需要主时间列。连接器会自动追加一个名为 `timestamp
当缓存行数达到 `batchSize` 时会触发一次写入；写入器关闭时，也会把剩余缓存数据提交到 Druid。当前没有按时间间隔定期刷新的配置。
+该 Sink 适合追加式批量写入，不会把 CDC 的更新/删除行自动转换成 Druid 的 upsert 或 delete 操作。
+
## 示例
### 写入单表
diff --git a/docs/zh/connectors/sink/GraphQL.md
b/docs/zh/connectors/sink/GraphQL.md
index 0d23e35d55..9b16ee96f0 100644
--- a/docs/zh/connectors/sink/GraphQL.md
+++ b/docs/zh/connectors/sink/GraphQL.md
@@ -55,6 +55,7 @@ mutation 一起发送。
- 上游字段名应和 mutation 里使用的变量名保持一致。
- 如果 `variables` 已有某个 key，并且设置 `valueCover = true`，会保留配置里的变量值。
- 如果 `valueCover = false`，同名输入行字段会覆盖配置里的变量值。
+- Sink 会为每一行输入数据发送一次 mutation 请求。如果上游包含多张表，需要确认同一条 mutation 和变量名可以处理所有写入到该
Sink 的表。
## 任务示例
@@ -154,6 +155,64 @@ sink {
}
```
+### 写入多张上游表
+
+```hocon
+source {
+ FakeSource {
+ plugin_output = "fake"
+ tables_configs = [
+ {
+ schema = {
+ table = "graphql_sink_1"
+ fields {
+ id = int
+ val_bool = boolean
+ val_string = string
+ }
+ }
+ rows = [
+ {
+ kind = INSERT
+ fields = [1, true, "NEW"]
+ }
+ ]
+ },
+ {
+ schema = {
+ table = "graphql_sink_2"
+ fields {
+ id = int
+ val_bool = boolean
+ val_string = string
+ }
+ }
+ rows = [
+ {
+ kind = INSERT
+ fields = [2, true, "READY"]
+ }
+ ]
+ }
+ ]
+ }
+}
+
+sink {
+ GraphQL {
+ plugin_input = "fake"
+ url = "http://graphql:8080/v1/graphql";
+ query = """
+ mutation MyMutation($id: Int!, $val_bool: Boolean!, $val_string:
String!) {
+ insert_sink(objects: {id: $id, val_bool: $val_bool, val_string:
$val_string}) {
+ affected_rows
+ }
+ }
+ """
+ }
+}
+```
+
## 变更日志
<ChangeLog />
diff --git a/docs/zh/connectors/sink/Hive.md b/docs/zh/connectors/sink/Hive.md
index ecd5667edd..49f4759cb1 100644
--- a/docs/zh/connectors/sink/Hive.md
+++ b/docs/zh/connectors/sink/Hive.md
@@ -100,6 +100,8 @@ Kerberos 的 keytab 文件路径
在中止操作期间是否从 Hive Metastore 中删除分区元数据的标志。注意：这只影响元存储中的元数据，分区中的数据将始终被删除（同步过程中生成的数据）。
+默认值为 `false`。
+
### parquet_avro_write_timestamp_as_int96 [boolean]
支持从时间戳写入 Parquet INT96，仅对 parquet 文件有效。
@@ -121,6 +123,8 @@ Kerberos 的 keytab 文件路径
注意：overwrite=true 与 data_save_mode=DROP_DATA 行为等价，二者择一配置即可，勿同时设置。
+批处理作业中，如果希望本次运行替换目标 Hive 表里的已有数据，可以使用 `overwrite = true` 或 `data_save_mode =
"DROP_DATA"`。普通追加写入场景保持默认的 `data_save_mode = "APPEND_DATA"` 即可。
+
### schema_save_mode [枚举]
在开始同步任务之前，针对目标端已存在的表结构选择不同的处理方案。
diff --git a/docs/zh/connectors/source/GraphQL.md
b/docs/zh/connectors/source/GraphQL.md
index 2b6a4a33b9..dbacccc1b9 100644
--- a/docs/zh/connectors/source/GraphQL.md
+++ b/docs/zh/connectors/source/GraphQL.md
@@ -64,6 +64,8 @@ GraphQL 源连接器用于从 GraphQL 服务读取数据。它支持：
- 订阅模式下，需要设置 `enable_subscription = true`，并使用 `ws://` 或 `wss://` 地址。
- Source 不支持 GraphQL `mutation` 操作。
- GraphQL 响应通常包在 `data` 字段下面，所以一般需要配置 `content_field`。
+- 使用 HTTP 流式轮询读取时，可以通过 `poll_interval_millis` 控制重复发送同一条查询的间隔。
+- 使用 WebSocket 订阅读取时，`max_retries` 和 `retry_delay_ms` 只控制订阅连接失败后的重连行为。
## 任务示例
