[
https://issues.apache.org/jira/browse/HBASE-11376?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14039600#comment-14039600
]
Enis Soztutar commented on HBASE-11376:
---------------------------------------
I think the problem is that LTT creates splits using the HexStringSplit which
is fine for LTT because it uses md5 prefixed row keys:
{code}
public byte[] getDeterministicUniqueKey(long keyBase) {
return LoadTestKVGenerator.md5PrefixedKey(keyBase).getBytes();
}
{code}
And HexStringSplit:
{code}
/**
* HexStringSplit is a well-known {@link SplitAlgorithm} for choosing region
* boundaries. The format of a HexStringSplit region boundary is the ASCII
* representation of an MD5 checksum, or any other uniformly distributed
* hexadecimal value. Row are hex-encoded long values in the range
* <b>"00000000" => "FFFFFFFF"</b> and are left-padded with zeros to keep the
* same order lexicographically as if they were binary.
*
* Since this split algorithm uses hex strings as keys, it is easy to read &
* write in the shell but takes up more space and may be non-intuitive.
*/
{code}
While in ITBLL, we want splits to be generated using UniformSplit rather than
HexStringSplit. I think it should be better if you create a patch which creates
the table pre-split using the UniformSplit, and not change the key format from
the generator.
{code}
/**
* A SplitAlgorithm that divides the space of possible keys evenly. Useful
* when the keys are approximately uniform random bytes (e.g. hashes). Rows
* are raw byte values in the range <b>00 => FF</b> and are right-padded with
* zeros to keep the same memcmp() order. This is the natural algorithm to use
* for a byte[] environment and saves space, but is not necessarily the
* easiest for readability.
*/
public static class UniformSplit implements SplitAlgorithm {
{code}
> IntegrationTestBigLinkedList's Generator tool does not generate keys
> belonging to all regions in a large table.
> ---------------------------------------------------------------------------------------------------------------
>
> Key: HBASE-11376
> URL: https://issues.apache.org/jira/browse/HBASE-11376
> Project: HBase
> Issue Type: Improvement
> Reporter: Vandana Ayyalasomayajula
> Assignee: Vandana Ayyalasomayajula
> Priority: Minor
> Attachments: HBASE-11376_1.patch, HBASE-11376_2.patch
>
>
> When IntegrationTestBigLinkedList's generator tool is used to generate keys
> to a large table ( 2200 regions), only some regions have keys and others are
> empty. It would be good to generate keys to all the regions of the table.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
