[HDFS-529] More redundant block searches in BlockManager.

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.21.0
• Fix Version/s: 0.21.0
• Component/s: namenode
• Hadoop Flags: Reviewed
• Assignee: Konstantin Shvachko
• Reporter: Konstantin Shvachko
• Created: 05/Aug/09 01:21
• Updated: 24/Aug/10 20:48
• Resolved: 06/Aug/09 02:40

Description

Found more redundant searches in BlockManager.addStoredBLock() and BlockManager.getNodes(Block).
The pattern here is that we first call BlocksMap.numNodes(Block) in order to determine the size of the array of data-nodes. This is one search in blocksMap. Then we call BlocksMap.nodeIterator(Block), which also performs a search. This is all not necessary because we already have the block in the form of BlockInfo from blocksMap and do not need to search for it.

Issue Links

• blocks: HDFS-517 Introduce BlockInfoUnderConstruction to reflect block replica states while writing.