Apache Hawq-使用中遇到的问题和解决方法

Posted 2020-12-30

df 查看函数列表
df+ 查看自定义函数

dc 查看转换
dc+

l 查看数据库列表

d table 查看表信息
d+ table 查看表详细信息

select * from information_schema.schemata;

1. HAWQ空值问题:
   HAWQ导入数据时，默认情况下空值必须使用"N"来表示

2.hawq load导入csv数据时，数据文件200G，启用一个gpfdist节点，花费时间1个小时40分钟，报错。
原因可能是内存不足。

drop table t1;
create table t1(id int, mac varchar(20), year int) 
with (appendonly=true,orientation=parquet,compresstype=snappy, bucketnum=18)
distributed by(mac)
PARTITION BY RANGE (year)
( START (2001)INCLUSIVE END (2003)EXCLUSIVE EVERY (1),
  DEFAULT PARTITION extra );

```

ALTER TABLE t1 SPLIT DEFAULT PARTITION
START (2003) INCLUSIVE
END (2004) EXCLUSIVE
INTO (PARTITION y2003, default partition);

报错：

NOTICE: exchanged partition "extra" of relation "t1" with relation "pg_temp_689110"
NOTICE: dropped partition "extra" for relation "t1"
ERROR: bucketnum requires a numeric value

解决方法： 去掉指定的bucketnum

set default_hash_table_bucket_number=9;
drop table t1;
create table t1(id int, mac varchar(20), year int)
with (appendonly=true,orientation=parquet,compresstype=snappy)
distributed by(mac)
PARTITION BY RANGE (year)
( START (2001)INCLUSIVE END (2003)EXCLUSIVE EVERY (1),
DEFAULT PARTITION extra );

ALTER TABLE t1 SPLIT DEFAULT PARTITION
START (2003) INCLUSIVE
END (2004) EXCLUSIVE
INTO (PARTITION y2003, default partition);
set default_hash_table_bucket_number=18;