[HDFS-14469] FsDatasetImpl() throws a DiskErrorException when the configuration has wrong values, results in unnecessary retry

• Type: Sub-task
• Status: Resolved
• Priority: Minor
• Resolution: Duplicate
• Assignee: Unassigned
• Reporter: Haicheng Chen
• Created: 06/May/19 13:55
• Updated: 29/May/19 18:07
• Resolved: 29/May/19 18:07

Description

Dear HDFS developers, we are developing a tool to detect exception-related bugs in Java. Our prototype has spotted the following throw statement whose exception class and error message indicate different error conditions.

 

Version: Hadoop-3.1.2

File: HADOOP-ROOT/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/fsdataset/impl/FsDatasetImpl.java

Line: 294-297

throw new DiskErrorException("Invalid value configured for "
    + "dfs.datanode.failed.volumes.tolerated - " + volFailuresTolerated
    + ". Value configured is either less than maxVolumeFailureLimit or greater than "
    + "to the number of configured volumes (" + volsConfigured + ").");

 

A DiskErrorException means an error has occurred when the process is interacting with the disk, e.g., in org.apache.hadoop.util.DiskChecker.checkDirInternal() we have the following code (lines 97-98):

throw new DiskErrorException("Cannot create directory: " + dir.toString());

However, the error message of the first exception indicates that dfs.datanode.failed.volumes.tolerated is configured incorrectly, which means there is nothing wrong with the disk (yet). This mismatch could be a problem. For example, the callers trying to handle other DiskErrorException may accidentally (and incorrectly) handle the configuration error.