[ https://issues.apache.org/jira/browse/HBASE-18199?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16044945#comment-16044945 ]
stack commented on HBASE-18199: ------------------------------- +1 on commit. Looks reasonable. Are we writing inline? We'll hold up even processing while this is going on? > Race in NettyRpcConnection may cause call stuck in > BufferCallBeforeInitHandler forever > -------------------------------------------------------------------------------------- > > Key: HBASE-18199 > URL: https://issues.apache.org/jira/browse/HBASE-18199 > Project: HBase > Issue Type: Bug > Components: IPC/RPC > Affects Versions: 3.0.0, 1.4.0, 2.0.0-alpha-1 > Reporter: Duo Zhang > Assignee: Duo Zhang > Fix For: 3.0.0, 1.4.0, 2.0.0-alpha-2 > > Attachments: HBASE-18199.patch > > > Found this when implementing the POC of HDFS-9924. > The problem is that channel.writeAndFlush will find the first outbound > handler before scheduling the actual write task into event loop, so it is > possible that when we actually execute the write task, the first outbound > handler has already been changed but the task still refers to the old handler > which has already been removed. -- This message was sent by Atlassian JIRA (v6.3.15#6346)