[HDFS-4181] LeaseManager tries to double remove and prints extra messages

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 0.23.4, 2.0.2-alpha
• Fix Version/s: 2.0.3-alpha, 0.23.5
• Component/s: namenode
• Target Version/s: 2.0.3-alpha, 0.23.5
• Hadoop Flags: Reviewed
• Assignee: Kihwal Lee
• Reporter: Kihwal Lee
• Created: 12/Nov/12 21:17
• Updated: 03/Sep/14 23:19
• Resolved: 13/Nov/12 15:34

Description

When checkLeases() runs, internalReleaseLease() is called on the expired ones. When it returns true, the lease is already removed, yet it is tried again in checkLease(). This causes unnecessary ERROR messages to be logged. The line doing removing.add(p) should be removed.

The internalReleaseLease() method logs a detailed message per call, so the extra INFO log message from checkLease() is redundant.

The error message from removeLease() can be very big and needs to be cut down. When the namenode itself is holding a lot of leases for block recovery, hitting this error is very expensive. In one instance, slow block recovery caused the namenode to hold more than 42K leases. The one log line in this case was over 4 MB. The dump of data structure should be only enabled in debug mode.