hadoop-yarn-api-python-client

Python client for Apache Hadoop® YARN API

Package documentation: python-client-for-hadoop-yarn-api.readthedocs.org

REST API documentation: hadoop.apache.org

Compatibility

Library is compatible with Apache Hadoop 3.2.1.

If u have version other than mentioned (or vendored variant like Hortonworks), certain APIs might be not working or have differences in implementation. If u plan to use certain API long-term, you might want to make sure its not in Alpha stage in documentation.

Installation

From PyPI

pip install yarn-api-client

From Anaconda (conda forge)

conda install -c conda-forge yarn-api-client

From source code

git clone https://github.com/toidi/hadoop-yarn-api-python-client.git
pushd hadoop-yarn-api-python-client
python setup.py install
popd

Enabling support for Kerberos/SPNEGO Security

1. First option - using requests_kerberos package

To avoid deployment issues on a non Kerberized environment, the requests_kerberos dependency is optional and needs to be explicit installed in order to enable access to YARN console protected by Kerberos/SPNEGO.

pip install requests_kerberos

From python code

from yarn_api_client.history_server import HistoryServer
from requests_kerberos import HTTPKerberosAuth
history_server = HistoryServer('https://127.0.0.2:5678', auth=HTTPKerberosAuth())

PS: You need to get valid kerberos ticket in systemwide kerberos cache before running your code, otherwise calls to kerberized environment won't go through (run kinit before proceeding to run code)

2. Second option - using gssapi package

If you want to avoid using terminal calls, you have to perform SPNEGO handshake to retrieve ticket yourself. Full API documentation: https://pythongssapi.github.io/python-gssapi/latest/

Usage

CLI interface

1. First way

bin/yarn_client --help

2. Alternative way

python -m yarn_api_client --help

Programmatic interface

from yarn_api_client import ApplicationMaster, HistoryServer, NodeManager, ResourceManager
am = ApplicationMaster('https://127.0.0.2:5678')
app_information = am.application_information('application_id')

Changelog

1.0.0 Release

• Major cleanup of API.
  • Address/port parameters have been replaced with complete endpoints (includes scheme [e.g., http or https]).
  • ResourceManager has been updated to take a list of endpoints for improved HA support.
  • ResourceManager, ApplicationMaster, HistoryServer and NodeManager have been updated with methods corresponding to the latest REST API.
• pytest support on Windows has been provided.
• Documentation has been updated.

NOTE: Applications using APIs relative to releases prior to 1.0 should pin their dependency on yarn-api-client to less than 1.0 and are encouraged to update to 1.0 as soon as possible.

0.3.7 Release

• Honor configured HTTP Policy when no address is provided - enabling using of HTTPS in these cases.

0.3.6 Release

• Extend ResourceManager to allow applications to determine resource availability prior to submission.

0.3.5 Release

• Hotfix release to fix internal signature mismatch

0.3.4 Release

• More flexible support for discovering Hadoop configuration including multiple Resource Managers when HA is configured
• Properly support YARN post response codes

0.3.3 Release

• Properly set Content-Type in PUT requests
• Check for HADOOP_CONF_DIR env variable

0.3.2 Release

• Make Kerberos/SPNEGO dependency optional

0.3.1 Release

• Fix cluster_application_kill API

0.3.0 Release

• Add support for YARN endpoints protected by Kerberos/SPNEGO
• Moved to requests package for REST API invocation
• Remove http_con property, as connections are now managed by requests package

0.2.5 Release

• Fixed History REST API

0.2.4 Release

• Added compatibility with HA enabled Resource Manager