[HADOOP-16018] DistCp won't reassemble chunks when blocks per chunk > 0

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.2.0, 2.9.2
• Fix Version/s: 3.0.4, 3.2.1, 2.9.3, 3.1.3
• Component/s: tools/distcp
• Assignee: Kai Xie
• Reporter: Kai Xie
• Created: 20/Dec/18 15:26
• Updated: 02/Oct/19 17:15
• Resolved: 27/Feb/19 22:08

Description

I was investigating why hadoop-distcp-2.9.2 won't reassemble chunks of the same file when blocks per chunk has been set > 0.

In the CopyCommitter::commitJob, this logic can prevent chunks from reassembling if blocks per chunk is equal to 0:

if (blocksPerChunk > 0) {
  concatFileChunks(conf);
}

Then in CopyCommitter's ctor, blocksPerChunk is initialised from the config:

blocksPerChunk = context.getConfiguration().getInt(
DistCpOptionSwitch.BLOCKS_PER_CHUNK.getConfigLabel(), 0);

 

But here the config key DistCpOptionSwitch.BLOCKS_PER_CHUNK.getConfigLabel() will always returns empty string because it is constructed without config label:

BLOCKS_PER_CHUNK("",
new Option("blocksperchunk", true, "If set to a positive value, files"
+ "with more blocks than this value will be split into chunks of "
+ "<blocksperchunk> blocks to be transferred in parallel, and "
+ "reassembled on the destination. By default, <blocksperchunk> is "
+ "0 and the files will be transmitted in their entirety without "
+ "splitting. This switch is only applicable when the source file "
+ "system implements getBlockLocations method and the target file "
+ "system implements concat method"))

As a result it will fall back to the default value 0 for blocksPerChunk, and prevent the chunks from reassembling.

 

 

Issue Links

• is caused by: HADOOP-15850 CopyCommitter#concatFileChunks should check that the blocks per chunk is not 0
• is related to: HADOOP-11794 Enable distcp to copy blocks in parallel
• links to: GitHub Pull Request #451
• links to: GitHub Pull Request #451