简介
PostgreSQL FDW是一种外部访问接口,它可以被用来访问存储在外部的数据,这些数据可以是外部的PG数据库,也可以mysql、ClickHouse等数据库。
ClickHouse是一款快速的开源OLAP数据库管理系统,它是面向列的,允许使用SQL查询实时生成分析报告。
clickhouse_fdw是一个开源的外部数据包装器(FDW)用于访问ClickHouse列存数据库。
目前有以下两款clickhouse_fdw:
https://github.com/adjust/clickhouse_fdw
一直持续不断的有提交,目前支持PostgreSQL 11-13
https://github.com/Percona-Lab/clickhousedb_fdw
之前有一年时间没有动静,最近一段时间刚从adjust/clickhouse_fdw merge了一下,目前也支持PostgreSQL 11-13。
本文就以adjust/clickhouse_fdw为例。
安装
# libcurl >= 7.43.0 yum install libcurl-devel libuuid-devel git clone https://github.com/adjust/clickhouse_fdw.git cd clickhouse_fdw mkdir build && cd build cmake .. make && make install
使用
CH端:
生成测试表及数据,这里我们使用CH官网提供的Star Schema Benchmark
https://clickhouse.tech/docs/en/getting-started/example-datasets/star-schema/#star-schema-benchmark
模拟数据量:5张数据表,数据主要集中在lineorder*表,单表9000w rows左右、22G存储。
[root@vm101 ansible]# clickhouse client ClickHouse client version 20.8.9.6. Connecting to localhost:9000 as user default. Connected to ClickHouse server version 20.8.9 revision 54438. vm101 :) show tables; SHOW TABLES ┌─name───────────┐ │ customer │ │ lineorder │ │ lineorder_flat │ │ part │ │ supplier │ └────────────────┘ 5 rows in set. Elapsed: 0.004 sec. vm101 :) select count(*) from lineorder_flat; SELECT count(*) FROM lineorder_flat ┌──count()─┐ │ 89987373 │ └──────────┘ 1 rows in set. Elapsed: 0.005 sec. [root@vm101 ansible]# du -sh /clickhouse/data/default/lineorder_flat/ 22G /clickhouse/data/default/lineorder_flat/
PG端:
创建FDW插件
postgres=# create extension clickhouse_fdw ; CREATE EXTENSION postgres=# \dew List of foreign-data wrappers Name | Owner | Handler | Validator ----------------+----------+--------------------------+---------------------------- clickhouse_fdw | postgres | clickhousedb_fdw_handler | clickhousedb_fdw_validator (1 row)
创建CH外部服务器
postgres=# CREATE SERVER clickhouse_svr FOREIGN DATA WRAPPER clickhouse_fdw OPTIONS(host '10.0.0.101', port '9000', dbname 'default', driver 'binary'); CREATE SERVER postgres=# \des List of foreign servers Name | Owner | Foreign-data wrapper ----------------+----------+---------------------- clickhouse_svr | postgres | clickhouse_fdw (1 row)
创建用户映射
postgres=# CREATE USER MAPPING FOR CURRENT_USER SERVER clickhouse_svr OPTIONS (user 'default', password ''); CREATE USER MAPPING postgres=# \deu List of user mappings Server | User name ----------------+----------- clickhouse_svr | postgres (1 row)
创建外部表
postgres=# IMPORT FOREIGN SCHEMA "default" FROM SERVER clickhouse_svr INTO public; IMPORT FOREIGN SCHEMA postgres=# \det List of foreign tables Schema | Table | Server --------+----------------+---------------- public | customer | clickhouse_svr public | lineorder | clickhouse_svr public | lineorder_flat | clickhouse_svr public | part | clickhouse_svr public | supplier | clickhouse_svr (5 rows)
查询
postgres=# select count(*) from lineorder_flat ; count ---------- 89987373 (1 row) postgres=# select "LO_ORDERKEY","C_NAME" from lineorder_flat limit 5; LO_ORDERKEY | C_NAME -------------+-------------------- 3271 | Customer#000099173 3271 | Customer#000099173 3271 | Customer#000099173 3271 | Customer#000099173 5607 | Customer#000273061 (5 rows)
需要注意的是CH是区分大小写的以及一些函数兼容问题,上面的示例也有展示。
测试SQL直接使用CH SSB提供的13条SQL,SQL基本类似,选一条做下测试,运行时间基本是一致的。
CH:
vm101 :) SELECT :-] toYear(LO_ORDERDATE) AS year, :-] C_NATION, :-] sum(LO_REVENUE - LO_SUPPLYCOST) AS profit :-] FROM lineorder_flat :-] WHERE C_REGION = 'AMERICA' AND S_REGION = 'AMERICA' AND (P_MFGR = 'MFGR#1' OR P_MFGR = 'MFGR#2') :-] GROUP BY :-] year, :-] C_NATION :-] ORDER BY :-] year ASC, :-] C_NATION ASC; SELECT toYear(LO_ORDERDATE) AS year, C_NATION, sum(LO_REVENUE - LO_SUPPLYCOST) AS profit FROM lineorder_flat WHERE (C_REGION = 'AMERICA') AND (S_REGION = 'AMERICA') AND ((P_MFGR = 'MFGR#1') OR (P_MFGR = 'MFGR#2')) GROUP BY year, C_NATION ORDER BY year ASC, C_NATION ASC ┌─year─┬─C_NATION──────┬───────profit─┐ │ 1992 │ ARGENTINA │ 157402521853 │ ... │ 1998 │ UNITED STATES │ 89854580268 │ └──────┴───────────────┴──────────────┘ 35 rows in set. Elapsed: 0.195 sec. Processed 89.99 million rows, 1.26 GB (460.70 million rows/s., 6.46 GB/s.)
PG:
postgres=# SELECT date_part('year', "LO_ORDERDATE") AS year, "C_NATION", sum("LO_REVENUE" - "LO_SUPPLYCOST") AS profit FROM lineorder_flat WHERE "C_REGION" = 'AMERICA' AND "S_REGION" = 'AMERICA' AND ("P_MFGR" = 'MFGR#1' OR "P_MFGR" = 'MFGR#2') GROUP BY year, "C_NATION" ORDER BY year ASC, "C_NATION" ASC; year | C_NATION | profit ------+---------------+-------------- 1992 | ARGENTINA | 157402521853 ... 1998 | UNITED STATES | 89854580268 (35 rows) Time: 195.102 ms