[ARROW-2025] [Python/C++] HDFS Client disconnect closes all open clients

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.9.0
• Component/s: C++, Python
• Labels: pull-request-available
• External issue URL: https://github.com/apache/arrow/issues/18006
• Assignee: Jim Crist
• Reporter: Jim Crist
• Created: 24/Jan/18 17:14
• Updated: 11/Jan/23 07:18
• Resolved: 25/Jan/18 01:33

Description

In the python library, if an instance of `HadoopFileSystem` is garbage collected, all other existing instances become invalid. I haven't checked with a C++ only example, but from reading the cython code I can't see how cython is responsible, so I think this is a bug in the C++ library.

 

>>> import pyarrow as pa
>>> h = pa.hdfs.connect()
18/01/24 16:54:25 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
18/01/24 16:54:26 WARN shortcircuit.DomainSocketFactory: The short-circuit local reads feature cannot be used because libhadoop cannot be loaded.
>>> h.ls("/")
['/benchmarks', '/hbase', '/tmp', '/user', '/var']
>>> h2 = pa.hdfs.connect()
>>> del h  # close one client
>>> h2.ls("/")  # all filesystem operations now fail
hdfsListDirectory(/): FileSystem#listStatus error:
IOException: Filesystem closedjava.io.IOException: Filesystem closed
        at org.apache.hadoop.hdfs.DFSClient.checkOpen(DFSClient.java:865)
        at org.apache.hadoop.hdfs.DFSClient.listPaths(DFSClient.java:2106)
        at org.apache.hadoop.hdfs.DFSClient.listPaths(DFSClient.java:2092)
        at org.apache.hadoop.hdfs.DistributedFileSystem.listStatusInternal(DistributedFileSystem.java:743)
        at org.apache.hadoop.hdfs.DistributedFileSystem.access$600(DistributedFileSystem.java:113)
        at org.apache.hadoop.hdfs.DistributedFileSystem$16.doCall(DistributedFileSystem.java:808)
        at org.apache.hadoop.hdfs.DistributedFileSystem$16.doCall(DistributedFileSystem.java:804)
        at org.apache.hadoop.fs.FileSystemLinkResolver.resolve(FileSystemLinkResolver.java:81)
        at org.apache.hadoop.hdfs.DistributedFileSystem.listStatus(DistributedFileSystem.java:804)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/opt/conda/lib/python3.6/site-packages/pyarrow/hdfs.py", line 88, in ls
    return super(HadoopFileSystem, self).ls(path, detail)
  File "io-hdfs.pxi", line 248, in pyarrow.lib.HadoopFileSystem.ls
  File "error.pxi", line 79, in pyarrow.lib.check_status
pyarrow.lib.ArrowIOError: HDFS: list directory failed
>>> h2.is_open  # The python object still thinks it's open
True

Issue Links

• links to: GitHub Pull Request #1499