[HDFS-14074] DataNode runs async disk checks maybe throws NullPointerException, and DataNode failed to register to NameSpace.

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.8.0, 3.0.0
• Fix Version/s: 3.3.0, 3.2.1, 3.1.3
• Component/s: hdfs
• Labels: HDFS HDFS-11114

Environment

hadoop-2.7.3, hadoop-2.8.0

• Target Version/s: 2.8.0, 2.7.3
• Hadoop Flags: Reviewed
• Tags: HDFS, DiskChecks
• Flags: Patch
• Assignee: guangyi lu
• Reporter: guangyi lu
• Created: 13/Nov/18 12:37
• Updated: 30/Mar/20 07:15
• Resolved: 22/Jun/19 01:29

Description

In ThrottledAsyncChecker class，it members of the completedChecks is WeakHashMap, its definition is as follows：
     this.completedChecks = new WeakHashMap<>();

and one of its uses is as follows in schedule method:
    if (completedChecks.containsKey(target)) 

{         // here may be happen garbage collection，and result may be null.        final LastCheckResult<V> result = completedChecks.get(target);                  final long msSinceLastCheck = timer.monotonicNow() - result.completedAt;            。。。。 }

after  "completedChecks.containsKey(target)"，  may be happen garbage collection，  and result may be null.

the solution is：

this.completedChecks = new ReferenceMap(1, 1);
or
this.completedChecks = new HashMap<>();

 

Issue Links

• is duplicated by: HDFS-14071 DataNode runs async disk checks maybe throws NullPointerException In ThrottledAsyncChecker.java
• is duplicated by: HDFS-14072 DataNode runs async disk checks maybe throws NullPointerException, and DataNode failed to register to NameSpace.
• is duplicated by: HDFS-14073 DataNode runs async disk checks maybe throws NullPointerException, and DataNode failed to register to NameSpace.
• is duplicated by: HDFS-14503 ThrottledAsyncChecker throws NPE during block pool initialization