How to download a file from HDFS

Downloading from HDFS should prove both that the source path exists and that the local copy is readable. The hdfs dfs -get and hadoop fs -copyToLocal commands copy files to the local filesystem used by the client shell.

Choose a destination path that will not overwrite local data unless replacement is intended. The -f flag overwrites existing files, while -crc writes checksum sidecar files for workflows that need transfer evidence.

Use HDFS permissions and Kerberos identity from the current shell. A command that works for one user can fail for another even on the same client host.

1. List the HDFS file before downloading it.

   $ hdfs dfs -ls /user/alice/input/events.csv
   -rw-r--r--   3 alice analytics   44040192 2026-06-17 03:14 /user/alice/input/events.csv

2. Download the file to the current directory.

   $ hdfs dfs -get /user/alice/input/events.csv ./events.csv

3. Check the local file size.

   $ ls -lh events.csv
   -rw-r--r--  1 alice  staff   42M Jun 17 03:42 events.csv

4. Compare the HDFS checksum when the filesystem supports it.

   $ hdfs dfs -checksum /user/alice/input/events.csv
   /user/alice/input/events.csv	MD5-of-0MD5-of-512CRC32C	0000020000000000000000007fb2c3a4

5. Read a small portion through the local tool that will consume the file.

   $ wc -l events.csv
   125000 events.csv