[HBASE-21699] Create table failed when using SPLITS_FILE => 'splits.txt'

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 2.0.0, 2.0.1, 2.1.1, 2.0.2, 2.0.3, 2.1.2, 2.0.4
• Fix Version/s: 3.0.0-alpha-1, 2.2.0, 2.1.3, 2.0.5, 2.3.0
• Component/s: Client, shell
• Flags: Patch
• Assignee: huan
• Reporter: huan
• Created: 09/Jan/19 03:26
• Updated: 30/Jan/19 19:40
• Resolved: 30/Jan/19 08:31

Description

Hi all:

 When I ran 

create 't1', 'f1', SPLITS_FILE => 'splits.txt'

on HBase2.0.0, it failed, and no detailed error info, just like below:

ERROR: 

Creates a table. Pass a table name, and a set of column family
specifications (at least one), and, optionally, table configuration.
Column specification can be a simple string (name), or a dictionary
(dictionaries are described below in main help output), necessarily
including NAME attribute.
Examples:

So I opened the debug:

hbase shell -d

and 

ERROR: 
Backtrace: org.apache.hadoop.hbase.util.Bytes.toBytes(org/apache/hadoop/hbase/util/Bytes.java:732)
org.apache.hadoop.hbase.HTableDescriptor.setValue(org/apache/hadoop/hbase/HTableDescriptor.java:190)

But it works on branch 1.2.0.

so I view the source code, I find the issue is because the below code:

// admin.rb
if arg.key?(SPLITS_FILE)
  splits_file = arg.delete(SPLITS_FILE)
  unless File.exist?(splits_file)
    raise(ArgumentError, "Splits file #{splits_file} doesn't exist")
  end
  arg[SPLITS] = []
  File.foreach(splits_file) do |line|
    arg[SPLITS].push(line.chomp)
  end
  htd.setValue(SPLITS_FILE, arg[SPLITS_FILE])
end

// HTableDescriptor part
public HTableDescriptor setValue(String key, String value) {
  getDelegateeForModification().setValue(Bytes.toBytes(key), Bytes.toBytes(value));
  return this;
}

// Bytes part
public static byte[] toBytes(String s) {
  try {
    return s.getBytes(UTF8_CSN);
  } catch (UnsupportedEncodingException e) {
    // should never happen!
    throw new IllegalArgumentException("UTF8 decoding is not supported", e);
  }
}

Call flow is:

admin.rb ---> htd.setValue(SPLITS_FILE, arg[SPLITS_FILE]) ---> Bytes.toBytes(key) && Bytes.toBytes(value) 

from Bytes.toBytes, if s is null, the function will throw NullPointerException, but HTableDescriptor.setValue(String key, String value) does not check key and value.

in admin.rb, it use arg.delete(SPLITS_FILE) to get the value, but this means, after using arg.delete(SPLITS_FILE),  arg[SPLITS_FILE] will return nil. so HTableDescriptor.setValue(String key, String value) does not check key and value is root cause.

why branch below 2.0.0 works fine, because old code is :

public HTableDescriptor setValue(String key, String value) {
if (value == null) {
remove(key);
} else {
setValue(Bytes.toBytes(key), Bytes.toBytes(value));
}
return this;
}

it check the value.

 since branch 2.0.0, HBase add new function called 'TableDescriptorBuilder'

it included:

public ModifyableTableDescriptor setValue(String key, String value) {
  return setValue(toBytesOrNull(key, Bytes::toBytes),
          toBytesOrNull(value, Bytes::toBytes));
}

it checked key and value, but HTableDescriptor.setValue(String key, String value)  does not call it, so

just change HTableDescriptor.setValue(String key, String value) to call it, it will works