[HDFS-4942] Add retry cache support in Namenode

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 2.1.0-beta
• Component/s: ha, namenode
• Assignee: Suresh Srinivas
• Reporter: Suresh Srinivas
• Created: 28/Jun/13 21:21
• Updated: 28/Sep/15 20:51
• Resolved: 20/May/15 04:11

Description

In current HA mechanism with FailoverProxyProvider and non HA setups with RetryProxy retry a request from the RPC layer. If the retried request has already been processed at the namenode, the subsequent attempts fail for non-idempotent operations such as create, append, delete, rename etc. This will cause application failures during HA failover, network issues etc.

This jira proposes adding retry cache at the namenode to handle these failures. More details in the comments.

Issue Links

• incorporates: HADOOP-9688 Add globally unique Client ID to RPC requests
• incorporates: HADOOP-9716 Move the Rpc request call ID generation to client side InvocationHandler
• incorporates: HADOOP-9717 Add retry attempt count to the RPC requests
• incorporates: HADOOP-9691 RPC clients can generate call ID using AtomicInteger instead of synchronizing on the Client instance.
• is related to: HDFS-4849 Enable retries for create and append operations.
• relates to: HADOOP-9786 RetryInvocationHandler#isRpcInvocation should support ProtocolTranslator
• relates to: HDFS-5008 Make ClientProtocol#abandonBlock() idempotent

Sub-Tasks

• Analyze and add annotations to Namenode protocol methods and enable retry: Closed
• Implement retry cache on the namenode: Closed
• Record ClientId and CallId in EditLog to enable rebuilding retry cache in case of HA failover: Closed
• Add metrics about the NameNode retry cache: Closed