Apache Phoenix非法数据异常

问题描述

我在从HBase写入数据并用Phoenix读取数据时遇到问题.这些是重现该问题的步骤:

I am having problems writing data from HBase and reading it with Phoenix.These are the steps to reproduce the problem:

使用Phoenix创建表.

Create a table using Phoenix.

 CREATE TABLE test (
    id varchar not null,
    t1.a unsigned_int,
    t1.b varchar
    CONSTRAINT pk PRIMARY KEY (id))
 COLUMN_ENCODED_BYTES = 0;

如果我使用Upsert使用Phoenix将信息添加到表中

If I add information to the table using Phoenix using Upsert

upsert into test (id, t1.a, t1.b) values ('a1',1,'foo_a');

然后我尝试查询表，我得到了:

And I try query the table, I get this:

select * from test;

+-----+----+--------+
| ID  | A  |   B    |
+-----+----+--------+
| a1  | 1  | foo_a  |
+-----+----+--------+

这时一切正常，但是现在我将直接使用HBase添加一个新条目.

At this point everything work as expected, but now I am going to add a new entry using HBase directly.

put 'TEST', 'id_1','T1:A', 2
put 'TEST', 'id_1','T1:B','some text'

此后，我无法再查询该表，得到此信息:

After that I can't query the table anymore, getting this:

select * from test;
Error: ERROR 201 (22000): Illegal data. Expected length of at least 4 bytes, but had 1 (state=22000,code=201)

我知道问题与HBase如何存储unsigned_int有关，如果我从表中删除此列，查询将再次起作用.该问题如何解决?

I know that the problem is related to how HBase is storing the unsigned_int, and if I remove this column from the table, the queries will work again.How can this problem be solved?

推荐答案

问题似乎与HBase如何存储数据有关，如果我对表进行了扫描，就会发现:

The problem seems to be related with how HBase is storing the data, if I make a scan of the table I get this:

ROW                           COLUMN+CELL
 a1                           column=T1:A, timestamp=1551274930165, value=\x00\x00\x00\x01
 a1                           column=T1:B, timestamp=1551274930165, value=foo_a
 a1                           column=T1:_0, timestamp=1551274930165, value=x
 id_1                         column=T1:A, timestamp=1551274993067, value=2
 id_1                         column=T1:B, timestamp=1551275070577, value=some text

这意味着新的整数值将作为字符串存储，因此正确的存储方式应为:

That means that the new integer value is being stored as a string, so the right way to store this data should be:

put 'TEST', 'id_1','T1:A', "\x00\x00\x00\x02"

完成此操作后，扫描将为我们提供:

Once this is done the scan will give us this:

ROW                           COLUMN+CELL
 a1                           column=T1:A, timestamp=1551274930165, value=\x00\x00\x00\x01
 a1                           column=T1:B, timestamp=1551274930165, value=foo_a
 a1                           column=T1:_0, timestamp=1551274930165, value=x
 id_1                         column=T1:A, timestamp=1551274993067, value=\x00\x00\x00\x02
 id_1                         column=T1:B, timestamp=1551275070577, value=some text

Phoenix可以毫无问题地访问数据.
感谢鲍里斯的提示.

And data will be accessible from Phoenix without any problem.
Thanks to Boris for the hint.

这篇关于Apache Phoenix非法数据异常的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！