Hudi系列7:使用SparkSQL操作Hudi

一. SparkSQL连接Hudi

1.1 Hive配置

我们需要将Hive 的 metastore服务独立出来

-- 目前只指定一个节点，也可以只用zookeeper做个高可用
cd $HIVE_HOME/conf
vi hive-site.xml
    
        hive.metastore.uris
        thrift://hp5:9083
    

然后启动hive metastore 服务

nohup hive --service metastore &
netstat -an | grep 9083

1.2 SparkSQL连接Hudi

# Spark 3.3
spark-sql --packages org.apache.hudi:hudi-spark3.3-bundle_2.12:0.12.0 \
--conf 'spark.serializer=org.apache.spark.serializer.KryoSerializer' \
--conf 'spark.sql.extensions=org.apache.spark.sql.hudi.HoodieSparkSessionExtension' \
--conf 'spark.sql.catalog.spark_catalog=org.apache.spark.sql.hudi.catalog.HoodieCatalog'

二. 创建表

创建表的时候有如下3个需要注意:

1. 表类型
   Hudi的两种表类型，即写时复制(COW)和读时合并(MOR)，都可以使用Spark SQL创建。在创建表时，可以使用type选项指定表的类型:type = 'cow’或type = ‘mor’。

2. 分区表和非分区表
   用户可以在Spark SQL中创建分区表或非分区表。要创建分区表，需要使用partitioned by语句指定分区列以创建分区表。当没有使用create table命令进行分区的语句时，该表被认为是一个非分区表。

3. Managed表和External表
   通常，Spark SQL支持两种表，即Managed表和External表。如果使用location语句或使用create external table显式地创建表来指定一个位置，则它是一个外部表，否则它被认为是一个托管表。你可以在这里阅读更多关于外部vs托管表的信息。

2.1 常规的建表

语法:
创建表的时候需要指定路径，不指定路径创建到本地了，Spark启用的是集群，其它节点访问不到，会产生报错。

-- 创建数据库
create database spark_hudi;
use spark_hudi;


--  创建一个表，不指定参数
create table hudi_cow_nonpcf_tbl (
  uuid int,
  name string,
  price double
) using hudi
location '/user/hudi/hudi_cow_nonpcf_tbl';


-- 创建一个MOR的非分区表
-- preCombineField 预聚合列 当id相同的时候，保留ts更大的那一条
create table hudi_mor_tbl (
  id int,
  name string,
  price double,
  ts bigint
) using hudi
tblproperties (
  type = 'mor',
  primaryKey = 'id',
  preCombineField = 'ts'
)
location '/user/hudi/hudi_mor_tbl';

-- 创建一个预聚合分区的COW表
create table hudi_cow_pt_tbl (
  id bigint,
  name string,
  ts bigint,
  dt string,
  hh string
) using hudi
tblproperties (
  type = 'cow',
  primaryKey = 'id',
  preCombineField = 'ts'
 )
partitioned by (dt, hh)
location '/user/hudi/hudi_cow_pt_tbl';

测试记录:

2.2 CTAS

代码:

-- CTAS: create a non-partitioned cow table without preCombineField
create table hudi_ctas_cow_nonpcf_tbl
using hudi
tblproperties (primaryKey = 'id')
location '/user/hudi/hudi_ctas_cow_nonpcf_tbl'
as
select 1 as id, 'a1' as name, 10 as price;


create table hudi_ctas_cow_pt_tbl
using hudi
tblproperties (type = 'cow', primaryKey = 'id', preCombineField = 'ts')
partitioned by (dt)
location '/user/hudi/hudi_ctas_cow_pt_tbl'
as
select 1 as id, 'a1' as name, 10 as price, 1000 as ts, '2021-12-01' as dt;

测试记录:
虽然建表过程看到有报错，但是依旧是成功的

三. 插入数据

-- insert into non-partitioned table
insert into hudi_cow_nonpcf_tbl select 1, 'a1', 20;
insert into hudi_mor_tbl select 1, 'a1', 20, 1000;

-- insert dynamic partition
insert into hudi_cow_pt_tbl partition (dt, hh)
select 1 as id, 'a1' as name, 1000 as ts, '2021-12-09' as dt, '10' as hh;

-- insert static partition
insert into hudi_cow_pt_tbl partition(dt = '2021-12-09', hh='11') select 2, 'a2', 1000;

-- upsert mode for preCombineField-provided table
insert into hudi_mor_tbl select 1, 'a1_1', 20, 1001;
select id, name, price, ts from hudi_mor_tbl;
1   a1_1    20.0    1001

-- bulk_insert mode for preCombineField-provided table
set hoodie.sql.bulk.insert.enable=true;
set hoodie.sql.insert.mode=non-strict;

insert into hudi_mor_tbl select 1, 'a1_2', 20, 1002;
select id, name, price, ts from hudi_mor_tbl;
1   a1_1    20.0    1001
1   a1_2    20.0    1002

测试记录:
虽然比insert hive_table快一些，但是感觉速度依旧不行

四. 查询数据

代码:

# 普通查询
 select fare, begin_lon, begin_lat, ts from  hudi_trips_snapshot where fare > 20.0;

# 基于时间线查询
create table hudi_cow_pt_tbl (
  id bigint,
  name string,
  ts bigint,
  dt string,
  hh string
) using hudi
tblproperties (
  type = 'cow',
  primaryKey = 'id',
  preCombineField = 'ts'
 )
partitioned by (dt, hh)
location '/user/hudi/hudi_cow_pt_tbl';

insert into hudi_cow_pt_tbl select 1, 'a0', 1000, '2021-12-09', '10';
select * from hudi_cow_pt_tbl;

-- record id=1 changes `name`
insert into hudi_cow_pt_tbl select 1, 'a1', 1001, '2021-12-09', '10';
select * from hudi_cow_pt_tbl;

-- time travel based on first commit time, assume `20221118154519532`
select * from hudi_cow_pt_tbl timestamp as of '20221118154519532' where id = 1;
-- time travel based on different timestamp formats
select * from hudi_cow_pt_tbl timestamp as of '2022-11-18 15:45:19.532' where id = 1;
select * from hudi_cow_pt_tbl timestamp as of '2022-03-08' where id = 1;

五. 更新数据

5.1 普通

语法:

UPDATE tableIdentifier SET column = EXPRESSION(,column = EXPRESSION) [ WHERE boolExpression]

代码:

update hudi_mor_tbl set price = price * 2, ts = 1111 where id = 1;

update hudi_cow_pt_tbl set name = 'a1_1', ts = 1001 where id = 1;

-- update using non-PK field
update hudi_cow_pt_tbl set ts = 1001 where name = 'a1';

5.2 MergeInto

语法:

MERGE INTO tableIdentifier AS target_alias
USING (sub_query | tableIdentifier) AS source_alias
ON 
[ WHEN MATCHED [ AND  ] THEN  ]
[ WHEN MATCHED [ AND  ] THEN  ]
[ WHEN NOT MATCHED [ AND  ]  THEN  ]

 =A equal bool condition 
  =
  DELETE  |
  UPDATE SET *  |
  UPDATE SET column1 = expression1 [, column2 = expression2 ...]
  =
  INSERT *  |
  INSERT (column1 [, column2 ...]) VALUES (value1 [, value2 ...])

代码:

-- source table using hudi for testing merging into non-partitioned table
create table hudi_merge_source (id int, name string, price double, ts bigint) using hudi
tblproperties (primaryKey = 'id', preCombineField = 'ts')
location '/user/hudi/hudi_merge_source';
insert into hudi_merge_source values (1, "old_a1", 22.22, 900), (2, "old_a2", 33.33, 2000), (3, "old_a3", 44.44, 2000);


create table hudi_merge_source2 (id int, name string, price double, ts bigint) using hudi
tblproperties (primaryKey = 'id', preCombineField = 'ts')
location '/user/hudi/hudi_merge_source2';
insert into hudi_merge_source2 values (2, "new_a2", 22.22, 900), (3, "new_a3", 33.33, 2000), (4, "new_a4", 44.44, 2000);


merge into hudi_merge_source as target
using (
  select * from hudi_merge_source2
) source
on target.id = source.id
when matched then
 update set name = source.name, price = source.price, ts = source.ts
when not matched then
 insert (id, name, price, ts) values(source.id, source.name, source.price, source.ts);

测试记录:

spark-sql> 
         > create table hudi_merge_source (id int, name string, price double, ts bigint) using hudi
         > tblproperties (primaryKey = 'id', preCombineField = 'ts')
         > location '/user/hudi/hudi_merge_source';
22/11/25 11:33:55 WARN DFSPropertiesConfiguration: Cannot find HUDI_CONF_DIR, please set it as the dir of hudi-defaults.conf
22/11/25 11:33:55 WARN DFSPropertiesConfiguration: Properties file file:/etc/hudi/conf/hudi-defaults.conf not found. Ignoring to load props file
22/11/25 11:33:58 WARN SessionState: METASTORE_FILTER_HOOK will be ignored, since hive.security.authorization.manager is set to instance of HiveAuthorizerFactory.
Time taken: 3.65 seconds
spark-sql> insert into hudi_merge_source values (1, "old_a1", 22.22, 900), (2, "old_a2", 33.33, 2000), (3, "old_a3", 44.44, 2000);
00:27  WARN: Timeline-server-based markers are not supported for HDFS: base path hdfs://hp5:8020/user/hudi/hudi_merge_source.  Falling back to direct markers.
00:32  WARN: Timeline-server-based markers are not supported for HDFS: base path hdfs://hp5:8020/user/hudi/hudi_merge_source.  Falling back to direct markers.
Time taken: 25.452 seconds
spark-sql> create table hudi_merge_source2 (id int, name string, price double, ts bigint) using hudi
         > tblproperties (primaryKey = 'id', preCombineField = 'ts')
         > location '/user/hudi/hudi_merge_source2';
Time taken: 0.541 seconds
spark-sql> insert into hudi_merge_source2 values (2, "new_a2", 22.22, 900), (3, "new_a3", 33.33, 2000), (4, "new_a4", 44.44, 2000);
00:58  WARN: Timeline-server-based markers are not supported for HDFS: base path hdfs://hp5:8020/user/hudi/hudi_merge_source2.  Falling back to direct markers.
01:02  WARN: Timeline-server-based markers are not supported for HDFS: base path hdfs://hp5:8020/user/hudi/hudi_merge_source2.  Falling back to direct markers.
Time taken: 11.574 seconds
spark-sql> merge into hudi_merge_source as target
         > using (
         >   select * from hudi_merge_source2
         > ) source
         > on target.id = source.id
         > when matched then
         >  update set name = source.name, price = source.price, ts = source.ts
         > when not matched then
         >  insert (id, name, price, ts) values(source.id, source.name, source.price, source.ts);
01:18  WARN: Timeline-server-based markers are not supported for HDFS: base path hdfs://hp5:8020/user/hudi/hudi_merge_source.  Falling back to direct markers.
01:21  WARN: Timeline-server-based markers are not supported for HDFS: base path hdfs://hp5:8020/user/hudi/hudi_merge_source.  Falling back to direct markers.
Time taken: 14.218 seconds
spark-sql> 
spark-sql> 
         > select * from hudi_merge_source2 ;
20221125113448990       20221125113448990_0_0   id:3            e3dec8f3-1c73-42dd-b1fa-b8d0c01748f5-0_0-64-2460_20221125113448990.parquet      3new_a3   33.33   2000
20221125113448990       20221125113448990_0_1   id:2            e3dec8f3-1c73-42dd-b1fa-b8d0c01748f5-0_0-64-2460_20221125113448990.parquet      2new_a2   22.22   900
20221125113448990       20221125113448990_0_2   id:4            e3dec8f3-1c73-42dd-b1fa-b8d0c01748f5-0_0-64-2460_20221125113448990.parquet      4new_a4   44.44   2000
Time taken: 0.781 seconds, Fetched 3 row(s)
spark-sql> select * from hudi_merge_source;
20221125113508944       20221125113508944_0_0   id:3            8ac8139e-0e9c-41f3-8046-24bf1b99aa9d-0_0-111-3707_20221125113508944.parquet     3new_a3   33.33   2000
20221125113412110       20221125113412110_0_1   id:1            8ac8139e-0e9c-41f3-8046-24bf1b99aa9d-0_0-111-3707_20221125113508944.parquet     1old_a1   22.22   900
20221125113412110       20221125113412110_0_2   id:2            8ac8139e-0e9c-41f3-8046-24bf1b99aa9d-0_0-111-3707_20221125113508944.parquet     2old_a2   33.33   2000
20221125113508944       20221125113508944_0_3   id:4            8ac8139e-0e9c-41f3-8046-24bf1b99aa9d-0_0-111-3707_20221125113508944.parquet     4new_a4   44.44   2000
Time taken: 1.231 seconds, Fetched 4 row(s)
spark-sql> 

六. 删除数据

Apache Hudi支持两种类型的删除:
(1)软删除:保留记录键，只清除所有其他字段的值(软删除中为空的记录始终保存在存储中，而不会删除);
(2)硬删除:从表中物理删除记录的任何痕迹。详细信息请参见写入数据页面的删除部分。

Spark SQL目前只支持硬删除

语法:

DELETE FROM tableIdentifier [ WHERE BOOL_EXPRESSION]

代码:

delete from hudi_merge_source where id = 1;

七. Insert Overwrite

代码:

-- insert overwrite non-partitioned table
insert overwrite hudi_mor_tbl select 99, 'a99', 20.0, 900;
insert overwrite hudi_cow_nonpcf_tbl select 99, 'a99', 20.0;

-- insert overwrite partitioned table with dynamic partition
insert overwrite table hudi_cow_pt_tbl select 10, 'a10', 1100, '2021-12-09', '10';

-- insert overwrite partitioned table with static partition
insert overwrite hudi_cow_pt_tbl partition(dt = '2021-12-09', hh='12') select 13, 'a13', 1100;

参考:

1. https://hudi.apache.org/docs/quick-start-guide/