[ https://issues.apache.org/jira/browse/HDFS-9442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15039546#comment-15039546 ]
Mingliang Liu commented on HDFS-9442: ------------------------------------- Failing tests can not re-produce locally. Specially, {{TestSeveralNameNodes}} is known to fail occasionally, tracked by [HDFS-9376]. {{TestBackupNode}} always binds to port 50070, which can cause bind failures here. It's tracked by [HDFS-9458]. > Move block replication logic from BlockManager to a new class > ReplicationManager > -------------------------------------------------------------------------------- > > Key: HDFS-9442 > URL: https://issues.apache.org/jira/browse/HDFS-9442 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: namenode > Reporter: Mingliang Liu > Assignee: Mingliang Liu > Attachments: HDFS-9442.000.patch, HDFS-9442.001.patch, > HDFS-9442.002.patch, HDFS-9442.003.patch, HDFS-9442.004.patch, > HDFS-9442.005.patch, HDFS-9442.006.patch > > > Currently the {{BlockManager}} is managing all replication logic for over- , > under- and mis-replicated blocks. This jira proposes to move that code to a > new class named {{ReplicationManager}} for cleaner code logic, shorter source > files, and easier lock separating work in future. > The {{ReplicationManager}} is a package local class, providing > {{BlockManager}} with methods that accesses its internal data structures of > replication queue. Meanwhile, the class maintains the lifecycle of > {{replicationThread}} and {{replicationQueuesInitializer}} daemon. -- This message was sent by Atlassian JIRA (v6.3.4#6332)