[HDFS-1407] Use Block in DataTransferProtocol

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.22.0
• Hadoop Flags: Reviewed
• Assignee: Suresh Srinivas
• Reporter: Suresh Srinivas
• Created: 18/Sep/10 00:06
• Updated: 12/Dec/11 06:19
• Resolved: 20/Sep/10 23:47

Description

Currently DataTransferProtocol has methods such as:

    public static void opReadBlock(DataOutputStream out, long blockId,
        long blockGs, long blockOffset, long blockLen, String clientName,
        Token<BlockTokenIdentifier> blockToken) throws IOException;

The client has to pass the individual elements that make block identification such as blockId and generation stamp. I propose methods with the following format:

    public static void opReadBlock(DataOutputStream out, Block block,
        long blockOffset, long blockLen, String clientName,
        Token<BlockTokenIdentifier> blockToken) throws IOException;

With this, the client need not understand the internals of Block. It receives Block over RPC and sends it in DataTransferProtocol. This helps in making Block opaque to the client.

Issue Links

• is part of: HDFS-1400 HDFS federation: Introduce block pool ID into DataTransferProtocol
• relates to: HDFS-1467 Append pipeline never succeeds with more than one replica