HDFS-CLI is an interactive command line shell that makes interacting with the Hadoop Distribted Filesystem (HDFS) simpler and more intuitive than the standard command-line tools that come with Hadoop. If you're familiar with OS X, Linux, or even Windows terminal/console-based applications, then you are likely familiar with features such as tab completion, command history, and ANSI formatting.
Download the release files to a temp location. As a root user, chmod +x the 3 shell script files then run the 'setup.sh'. This will create and install the hdfscli application to your path.
Try it out on a host with default configs:
hdfscli
To use an alternate HADOOP_CONF_DIR:
hdfscli --config /var/hadoop/dev-cfg
- Due to the complexities and requirements of connecting to an environment, I've removed the
ability to connect manually to an environment by simply using 'connect'. This option was there from the beginning, but as more and more features are added, I'm finding myself hacking
away at recreating the settings and controls enabled through the configurations available in
hdfs-site.xml
andcore-site.xml
. Therefore, the options-k for kerberos
and-a for auto connect
are no longer available. Unless specified via the--config
option, thehdfs-site.xml
andcore-site.xml
files in the default location of/etc/hadoop/conf
will be used to establish all of the environment variables needed to connect. You can have multiple directories with varioushdfs|core-site.xml
files in them and use the--config
option to enable connectivity to alternate hadoop environments.
-a
Auto Connect. Use --config
for alternate site files or nothing for the default /etc/hadoop/conf
.
-k
Kerberos Option
I noticed some pauses coming from inquiries into the Namenode JMX for nnstat
. Instead of requesting the
entire Namenode Jmx stack, now we target only the JmxBeans that we're interested in. This will help
the observed pauses and relieve the Namenode of some unnecessary work.
- Added new 'nnstat' function to collect Namenode JMX stats for long term analysis.
See NN Stat Feature
Also checkout how to auto fetch the stats via a script. Use this technique to run cron jobs to gather the stats.
Where -i stats
defines the initialization file. See Auto nnstat
hdfscli -i stats
- Added new 'lsp' function. Consider it an 'ls' PLUS.
- External Config Support (See Below)
- Supports NN HA (thru hdfs and core site files)
- Auto Config support using a default config directory.
- Setup Script to help deploy (bin/setup.sh)
- hdfscli shell script to launch (bin/hdfscli.sh)
- Support for initialization Script (-i <file>)
- Kerberos Support via default and --config option for hadoop site files.
- Added support for create/delete/rename Snapshot
- Initial Forked Release of P. Taylor Goetz.
- Update to 2.6.0 Hadoop Libraries
- Re-wrote Command Implementation to use FSShell as basis for issuing commands.
- Provide Context Feedback in command window to show local and remote context.
- Added several missing hdfs dfs commands that didn't exist earlier.
This project requires the artifacts from https://github.com/dstreev/stemshell , which is a forked enhancement that has added support for processing command line parameters and deals with quoted variables.
HDFS-CLI works much like a command-line ftp client: You first establish a connection to a remote HDFS filesystem, then manage local/remote files and transfers.
To start HDFS-CLI, run the following command:
java -jar hdfs-cli-full-bin.jar
Help for any command can be obtained by executing the help
command:
help pwd
Note that currently, documentation may be limited.
When working within a HDFS-CLI session, you manage both local (on your computer) and remote (HDFS) files. By convention, commands that apply to both local and remote filesystems are differentiated by prepending an l
character to the name to denote "local".
For example:
lls
lists local files in the local current working directory.
ls
lists remote files in the remote current working directory.
Every HDFS-CLI session keeps track of both the local and remote current working directories.
By default, hdfs-cli will use /etc/hadoop/conf
as the default location to search for
core-site.xml
and hdfs-site.xml
. If you want to use an alternate, use the --config
option when starting up hdfs-cli.
The --config
option takes 1 parameter, a local directory. This directory should contain hdfs-site.xml and core-site.xml files. When used, you'll automatically be connected to hdfs and changed to you're hdfs home directory.
Example Connection parameters.
# Use the hadoop files in the input directory to configure and connect to HDFS.
hdfscli --config ../mydir
This can be used in conjunction with the 'Startup' Init option below to run a set of commands automatically after the connection is made. The 'connect' option should NOT be used in the initialization script.
Using the option '-i ' when launching the CLI, it will run all the commands in the file.
The file needs to be location in the $HOME/.hdfs-cli directory. For example:
# If you're using the helper shell script
hdfscli -i test
# If you're using the java command
java -jar hdfs-cli-full-bin.jar -i test
Will initialize the session with the command(s) in $HOME/.hdfs-cli/test. One command per line.
The contents could be any set of valid commands that you would use in the cli. For example:
cd user/dstreev
Like 'ls', you can fetch many details about a file. But with this, you can also add information about the file that includes:
- Block Size
- Access Time
- Ratio of File Size to Block
- Datanode information for the files blocks (Host and Block Id)
Use help to get the options:
help lsp
usage: stats [OPTION ...] [ARGS ...]
Options:
-d,--maxDepth <maxDepth> Depth of Recursion (default 5), use '-1'
for unlimited
-f,--format <output-format> Comma separated list of one or more:
permissions_long,replication,user,group,siz
e,block_size,ratio,mod,access,path,datanode
_info (default all of the above)
-o,--output <output> Output File (HDFS) (default System.out)
When not argument is specified, it will use the current directory.
Examples:
# Using the default format, output a listing to the files in `/user/dstreev/perf` to `/tmp/test.out`
lsp -o /tmp/test.out /user/dstreev/perf
Output with the default format of:
permissions_long,replication,user,group,size,block_size,ratio,mod,access,path,datanode_info
rw-------,3,dstreev,hdfs,429496700,134217728,3.200,2015-10-24 12:26:39.689,2015-10-24 12:23:27.406,/user/dstreev/perf/teragen_27/part-m-00004,10.0.0.166,d2.hdp.local,blk_1073747900
rw-------,3,dstreev,hdfs,429496700,134217728,3.200,2015-10-24 12:26:39.689,2015-10-24 12:23:27.406,/user/dstreev/perf/teragen_27/part-m-00004,10.0.0.167,d3.hdp.local,blk_1073747900
rw-------,3,dstreev,hdfs,33,134217728,2.459E-7,2015-10-24 12:27:09.134,2015-10-24 12:27:06.560,/user/dstreev/perf/terasort_27/_partition.lst,10.0.0.166,d2.hdp.local,blk_1073747909
rw-------,3,dstreev,hdfs,33,134217728,2.459E-7,2015-10-24 12:27:09.134,2015-10-24 12:27:06.560,/user/dstreev/perf/terasort_27/_partition.lst,10.0.0.167,d3.hdp.local,blk_1073747909
rw-------,1,dstreev,hdfs,543201700,134217728,4.047,2015-10-24 12:29:28.706,2015-10-24 12:29:20.882,/user/dstreev/perf/terasort_27/part-r-00002,10.0.0.167,d3.hdp.local,blk_1073747920
rw-------,1,dstreev,hdfs,543201700,134217728,4.047,2015-10-24 12:29:28.706,2015-10-24 12:29:20.882,/user/dstreev/perf/terasort_27/part-r-00002,10.0.0.167,d3.hdp.local,blk_1073747921
With the file in HDFS, you can build a hive table on top of it to do some analysis. One of the reasons I created this was to be able to review a directory used by some process and get a baring on the file construction and distribution across the cluster.
- The ratio can be used to identify files that are below the block size (small files).
- With the Datanode information, you can determine if a dataset is hot-spotted on a cluster. All you need is a full list of hosts to join the results with.
help display help information
put upload local files to the remote HDFS
get (todo) retrieve remote files from HDFS to Local Filesystem
cd change current working directory
ls list directory contents
rm delete files/directories
pwd print working directory path
cat print file contents
chown change ownership
chmod change permissions
chgrp change group
head print first few lines of a file
mkdir create directories
count Count the number of directories, files and bytes under the paths that match the specified file pattern.
stat Print statistics about the file/directory at <path> in the specified format.
tail Displays last kilobyte of the file to stdout.
text Takes a source file and outputs the file in text format.
touchz Create a file of zero length.
usage Return the help for an individual command.
createSnapshot Create Snapshot
deleteSnapshot Delete Snapshot
renameSnapshot Rename Snapshot
lcd change current working directory
lls list directory contents
lrm delete files/directories
lpwd print working directory path
lcat print file contents
lhead print first few lines of a file
lmkdir create directories
lsp ls plus. Includes Block information and locations.
nnstat Namenode Statistics
- No support for paths containing spaces
- No support for Windows XP
- Path Completion for chown, chmod, chgrp, rm is broken
- Support input variables
- Expand to support Extended ACL's (get/set)
- Add Support for setrep
- HA Commands
- NN and RM