快速体验 Apache Doris & Iceberg

使用限制

  1. 支持 Iceberg V1/V2 表格式。
  2. 支持 Position Delete。
  3. 2.1.3 版本开始支持 Equality Delete。
  4. 支持 Parquet 文件格式
  5. 2.1.3 版本开始支持 ORC 文件格式。

创建 Catalog

基于 Hive Metastore 创建 Catalog

和 Hive Catalog 基本一致,这里仅给出简单示例。其他示例可参阅 Hive Catalog

  1. CREATE CATALOG iceberg PROPERTIES (
  2. 'type'='hms',
  3. 'hive.metastore.uris' = 'thrift://172.21.0.1:7004',
  4. 'hadoop.username' = 'hive',
  5. 'dfs.nameservices'='your-nameservice',
  6. 'dfs.ha.namenodes.your-nameservice'='nn1,nn2',
  7. 'dfs.namenode.rpc-address.your-nameservice.nn1'='172.21.0.2:4007',
  8. 'dfs.namenode.rpc-address.your-nameservice.nn2'='172.21.0.3:4007',
  9. 'dfs.client.failover.proxy.provider.your-nameservice'='org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider'
  10. );

基于 Iceberg API 创建 Catalog

使用 Iceberg API 访问元数据的方式,支持 Hadoop File System、Hive、REST、Glue、DLF 等服务作为 Iceberg 的 Catalog。

Hadoop Catalog

注意:warehouse 的路径必须指向 Database 路径的上一级。

示例:如果你的表路径是:s3://bucket/path/to/db1/table1,那么 warehouse 应该是:s3://bucket/path/to/

  1. CREATE CATALOG iceberg_hadoop PROPERTIES (
  2. 'type'='iceberg',
  3. 'iceberg.catalog.type' = 'hadoop',
  4. 'warehouse' = 'hdfs://your-host:8020/dir/key'
  5. );
  1. CREATE CATALOG iceberg_hadoop_ha PROPERTIES (
  2. 'type'='iceberg',
  3. 'iceberg.catalog.type' = 'hadoop',
  4. 'warehouse' = 'hdfs://your-nameservice/dir/key',
  5. 'dfs.nameservices'='your-nameservice',
  6. 'dfs.ha.namenodes.your-nameservice'='nn1,nn2',
  7. 'dfs.namenode.rpc-address.your-nameservice.nn1'='172.21.0.2:4007',
  8. 'dfs.namenode.rpc-address.your-nameservice.nn2'='172.21.0.3:4007',
  9. 'dfs.client.failover.proxy.provider.your-nameservice'='org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider'
  10. );
  1. CREATE CATALOG iceberg_s3 PROPERTIES (
  2. 'type'='iceberg',
  3. 'iceberg.catalog.type' = 'hadoop',
  4. 'warehouse' = 's3://bucket/dir/key',
  5. 's3.endpoint' = 's3.us-east-1.amazonaws.com',
  6. 's3.access_key' = 'ak',
  7. 's3.secret_key' = 'sk'
  8. );

Hive Metastore

  1. CREATE CATALOG iceberg PROPERTIES (
  2. 'type'='iceberg',
  3. 'iceberg.catalog.type'='hms',
  4. 'hive.metastore.uris' = 'thrift://172.21.0.1:7004',
  5. 'hadoop.username' = 'hive',
  6. 'dfs.nameservices'='your-nameservice',
  7. 'dfs.ha.namenodes.your-nameservice'='nn1,nn2',
  8. 'dfs.namenode.rpc-address.your-nameservice.nn1'='172.21.0.2:4007',
  9. 'dfs.namenode.rpc-address.your-nameservice.nn2'='172.21.0.3:4007',
  10. 'dfs.client.failover.proxy.provider.your-nameservice'='org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider'
  11. );

AWS Glue

连接 Glue 时,如果是在非 EC2 环境,需要将 EC2 环境里的 ~/.aws 目录拷贝到当前环境里。也可以下载AWS Cli工具进行配置,这种方式也会在当前用户目录下创建.aws目录。

  1. CREATE CATALOG glue PROPERTIES (
  2. "type"="iceberg",
  3. "iceberg.catalog.type" = "glue",
  4. "glue.endpoint" = "https://glue.us-east-1.amazonaws.com",
  5. "glue.access_key" = "ak",
  6. "glue.secret_key" = "sk"
  7. );
  1. Iceberg 属性详情参见 Iceberg Glue Catalog

  2. 如果在 AWS 服务(如 EC2)中,不填写 Credentials 相关信息(glue.access_keyglue.secret_key),Doris 就会使用默认的 DefaultAWSCredentialsProviderChain,它会读取系统环境变量或者 InstanceProfile 中配置的属性。

阿里云 DLF

参见阿里云 DLF Catalog 配置

REST Catalog

该方式需要预先提供 REST 服务,用户需实现获取 Iceberg 元数据的 REST 接口。

  1. CREATE CATALOG iceberg PROPERTIES (
  2. 'type'='iceberg',
  3. 'iceberg.catalog.type'='rest',
  4. 'uri' = 'http://172.21.0.1:8181'
  5. );

如果使用 HDFS 存储数据,并开启了高可用模式,还需在 Catalog 中增加 HDFS 高可用配置:

  1. CREATE CATALOG iceberg PROPERTIES (
  2. 'type'='iceberg',
  3. 'iceberg.catalog.type'='rest',
  4. 'uri' = 'http://172.21.0.1:8181',
  5. 'dfs.nameservices'='your-nameservice',
  6. 'dfs.ha.namenodes.your-nameservice'='nn1,nn2',
  7. 'dfs.namenode.rpc-address.your-nameservice.nn1'='172.21.0.1:8020',
  8. 'dfs.namenode.rpc-address.your-nameservice.nn2'='172.21.0.2:8020',
  9. 'dfs.client.failover.proxy.provider.your-nameservice'='org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider'
  10. );

Google Dataproc Metastore

  1. CREATE CATALOG iceberg PROPERTIES (
  2. "type"="iceberg",
  3. "iceberg.catalog.type"="hms",
  4. "hive.metastore.uris" = "thrift://172.21.0.1:9083",
  5. "gs.endpoint" = "https://storage.googleapis.com",
  6. "gs.region" = "us-east-1",
  7. "gs.access_key" = "ak",
  8. "gs.secret_key" = "sk",
  9. "use_path_style" = "true"
  10. );

hive.metastore.uris: Dataproc Metastore 服务开放的接口,在 Metastore 管理页面获取:Dataproc Metastore Services.

Iceberg On Object Storage

若数据存放在 S3 上,properties 中可以使用以下参数:

  1. "s3.access_key" = "ak"
  2. "s3.secret_key" = "sk"
  3. "s3.endpoint" = "s3.us-east-1.amazonaws.com"
  4. "s3.region" = "us-east-1"

数据存放在阿里云 OSS 上:

  1. "oss.access_key" = "ak"
  2. "oss.secret_key" = "sk"
  3. "oss.endpoint" = "oss-cn-beijing-internal.aliyuncs.com"
  4. "oss.region" = "oss-cn-beijing"

数据存放在腾讯云 COS 上:

  1. "cos.access_key" = "ak"
  2. "cos.secret_key" = "sk"
  3. "cos.endpoint" = "cos.ap-beijing.myqcloud.com"
  4. "cos.region" = "ap-beijing"

数据存放在华为云 OBS 上:

  1. "obs.access_key" = "ak"
  2. "obs.secret_key" = "sk"
  3. "obs.endpoint" = "obs.cn-north-4.myhuaweicloud.com"
  4. "obs.region" = "cn-north-4"

示例

  1. -- MinIO & Rest Catalog
  2. CREATE CATALOG `iceberg` PROPERTIES (
  3. "type" = "iceberg",
  4. "iceberg.catalog.type" = "rest",
  5. "uri" = "http://10.0.0.1:8181",
  6. "warehouse" = "s3://bucket",
  7. "token" = "token123456",
  8. "s3.access_key" = "ak",
  9. "s3.secret_key" = "sk",
  10. "s3.endpoint" = "http://10.0.0.1:9000",
  11. "s3.region" = "us-east-1"
  12. );

列类型映射

Iceberg TypeDoris Type
booleanboolean
intint
longbigint
floatfloat
doubledouble
decimal(p,s)decimal(p,s)
datedate
uuidstring
timestamp (Timestamp without timezone)datetime(6)
timestamptz (Timestamp with timezone)datetime(6)
stringstring
fixed(L)char(L)
binarystring
structstruct(2.1.3 版本开始支持)
mapmap(2.1.3 版本开始支持)
listarray
time不支持

Time Travel

支持读取 Iceberg 表指定的 Snapshot。

每一次对 iceberg 表的写操作都会产生一个新的快照。

默认情况下,读取请求只会读取最新版本的快照。

可以使用 FOR TIME AS OFFOR VERSION AS OF 语句,根据快照 ID 或者快照产生的时间读取历史版本的数据。示例如下:

SELECT * FROM iceberg_tbl FOR TIME AS OF "2022-10-07 17:20:37";

SELECT * FROM iceberg_tbl FOR VERSION AS OF 868895038966572;

另外,可以使用 iceberg_meta 表函数查询指定表的 snapshot 信息。