[HDFS-12679] Namenode stale memory Datanode gets Wipes out

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 2.8.1
• Component/s: namenode
• Assignee: Unassigned
• Reporter: Vivek Ghatala
• Created: 18/Oct/17 17:37
• Updated: 18/Oct/17 17:37

Description

Namenode stale mapping information about datanodes commands wiping out datanode.

LOG LINES:

2017-10-11 23:01:45,969 DEBUG org.apache.hadoop.hdfs.server.blockmanagement.DatanodeManager: remove datanode XXX.XX.XX.1:YYYYY

2017-10-11 23:01:45,969 DEBUG org.apache.hadoop.hdfs.server.blockmanagement.DatanodeManager: DatanodeManager.wipeDatanode(XXX.XX.XX.1:YYYYY): storage ##STORAGE_ID1## is removed from datanodeMap.

Scenario: Our environment uses shared storage. Whenever some datanode restarts, some other node comes up with that storage with different Ip address. It is possible that multiple datanodes restarts at same time.

Case: the node 1 serving some storage X is now serving storage Y. node 2 comes up and is serving storage X now. Namenode here commands storage Y to get wiped out.