[HDFS-15801] Backport HDFS-14582 to branch-2.10 (Failed to start DN with ArithmeticException when NULL checksum used)

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 2.10.2
• Labels: pull-request-available
• Assignee: Janus Chow
• Reporter: Janus Chow
• Created: 29/Jan/21 07:14
• Updated: 03/Feb/21 01:10
• Resolved: 01/Feb/21 23:44

Description

In HDFS-14582, the error message is more clear as follows:

Caused by: java.lang.ArithmeticException: / by zero
        at org.apache.hadoop.hdfs.server.datanode.fsdataset.impl.BlockPoolSlice.validateIntegrityAndSetLength(BlockPoolSlice.java:823)
        at 

But in branch-2.10.1, the exception message is omitted as follows:

2021-01-29 14:20:30,694 INFO  impl.FsDatasetImpl (FsVolumeList.java:run(204)) - Caught exception while adding replicas from /mnt/disk/0/hdfs/data/current. Will throw later.
java.io.IOException: Failed to start sub tasks to add replica in replica map :java.lang.ArithmeticExceptionjava.io.IOException: Failed to start sub tasks to add replica in replica map :java.lang.ArithmeticException at org.apache.hadoop.hdfs.server.datanode.fsdataset.impl.BlockPoolSlice.getVolumeMap(BlockPoolSlice.java:434) at org.apache.hadoop.hdfs.server.datanode.fsdataset.impl.FsVolumeImpl.getVolumeMap(FsVolumeImpl.java:930) at org.apache.hadoop.hdfs.server.datanode.fsdataset.impl.FsVolumeList$1.run(FsVolumeList.java:196)

The specific error message is omitted, causing it harder to find the root cause.

Issue Links

• links to: GitHub Pull Request #2659