PyStellarDB 0.12.0

Python interface to StellarDB

PyStellarDB

PyStellarDB is a Python API for executing Transwarp Exetended OpenCypher(TEoC) and Hive query. It could also generate a RDD object which could be used in PySpark. It is base on PyHive(https://github.com/dropbox/PyHive) and PySpark(https://github.com/apache/spark/)

PySpark RDD

We hack a way to generate RDD object using the same method in sc.parallelize(data). It could cause memory panic if the query returns a large amount of data.

Users could use a workaround if you do need huge data:

1. If you are querying a graph, refer to StellarDB manual of Chapter 4.4.5 to save the query data into a temporary table.

2. If you are querying a SQL table, save your query result into a temporary table.

3. Find the HDFS path of the temporary table generated in Step 1 or Step 2.

4. Use API like sc.newAPIHadoopFile() to generate RDD.

Usage

PLAIN Mode (No security is configured)

from pystellardb import stellar_hive

conn = stellar_hive.StellarConnection(host="localhost", port=10000, graph_name='pokemon')
cur = conn.cursor()
cur.execute('config query.lang cypher')
cur.execute('use graph pokemon')
cur.execute('match p = (a)-[f]->(b) return a,f,b limit 1')

print cur.fetchall()

LDAP Mode

from pystellardb import stellar_hive

conn = stellar_hive.StellarConnection(host="localhost", port=10000, username='hive', password='123456', auth='LDAP', graph_name='pokemon')
cur = conn.cursor()
cur.execute('config query.lang cypher')
cur.execute('use graph pokemon')
cur.execute('match p = (a)-[f]->(b) return a,f,b limit 1')

print cur.fetchall()

Kerberos Mode

# Make sure you have the correct realms infomation about the KDC server in /etc/krb5.conf
# Make sure you have the correct keytab file in your environment
# Run kinit command:
# In Linux: kinit -kt FILE_PATH_OF_KEYTABL PRINCIPAL_NAME
# In Mac: kinit -t FILE_PATH_OF_KEYTABL -f PRINCIPAL_NAME

from pystellardb import stellar_hive

conn = stellar_hive.StellarConnection(host="localhost", port=10000, kerberos_service_name='hive', auth='KERBEROS', graph_name='pokemon')
cur = conn.cursor()
cur.execute('config query.lang cypher')
cur.execute('use graph pokemon')
cur.execute('match p = (a)-[f]->(b) return a,f,b limit 1')

print cur.fetchall()

Execute Hive Query

from pystellardb import stellar_hive

# If `graph_name` parameter is None, it will execute a Hive query and return data just as PyHive does
conn = stellar_hive.StellarConnection(host="localhost", port=10000, database='default')
cur = conn.cursor()
cur.execute('SELECT * FROM default.abc limit 10')

Execute Graph Query and change to a PySpark RDD object

from pyspark import SparkContext
from pystellardb import stellar_hive

sc = SparkContext("local", "Demo App")

conn = stellar_hive.StellarConnection(host="localhost", port=10000, graph_name='pokemon')
cur = conn.cursor()
cur.execute('config query.lang cypher')
cur.execute('use graph pokemon')
cur.execute('match p = (a)-[f]->(b) return a,f,b limit 10')

rdd = cur.toRDD(sc)

def f(x): print(x)

rdd.map(lambda x: (x[0].toJSON(), x[1].toJSON(), x[2].toJSON())).foreach(f)

# Every line of this query is in format of Tuple(VertexObject, EdgeObject, VertexObject)
# Vertex and Edge object has a function of toJSON() which can print the object in JSON format

Execute Hive Query and change to a PySpark RDD object

from pyspark import SparkContext
from pystellardb import stellar_hive

sc = SparkContext("local", "Demo App")

conn = stellar_hive.StellarConnection(host="localhost", port=10000)
cur = conn.cursor()
cur.execute('select * from default_db.default_table limit 10')

rdd = cur.toRDD(sc)

def f(x): print(x)

rdd.foreach(f)

# Every line of this query is in format of Tuple(Column, Column, Column)

Dependencies

Required:

• Python 2.7+ / Python 3

System SASL

Ubuntu:

apt-get install libsasl2-dev libsasl2-2 libsasl2-modules-gssapi-mit
apt-get install python-dev gcc              #Update python and gcc if needed

RHEL/CentOS:

yum install cyrus-sasl-md5 cyrus-sasl-plain cyrus-sasl-gssapi cyrus-sasl-devel
yum install gcc-c++ python-devel.x86_64     #Update python and gcc if needed

# if pip3 install fails with a message like 'Can't connect to HTTPS URL because the SSL module is not available'
# you may need to update ssl & reinstall python

# 1. Download a higher version of openssl, e.g: https://www.openssl.org/source/openssl-1.1.1k.tar.gz
# 2. Install openssl: ./config && make && make install
# 3. Link openssl: echo /usr/local/lib64/ > /etc/ld.so.conf.d/openssl-1.1.1.conf
# 4. Update dynamic lib: ldconfig -v
# 5. Uninstall Python & Download a new Python source package
# 6. vim Modules/Setup, search '_socket socketmodule.c', uncomment
#    _socket socketmodule.c
#    SSL=/usr/local/ssl
#    _ssl _ssl.c \
#            -DUSE_SSL -I$(SSL)/include -I$(SSL)/include/openssl \
#            -L$(SSL)/lib -lssl -lcrypto
#
# 7. Install Python: ./configure && make && make install

Windows:

# There are 3 ways of installing sasl for python on windows
# 1. (recommended) Download a .whl version of sasl from https://www.lfd.uci.edu/~gohlke/pythonlibs/#sasl
# 2. (recommended) If using anaconda, use conda install sasl.
# 3. Install Microsoft Visual C++ 9.0/14.0 buildtools for python2.7/3.x, then pip install sasl.

Notices

Pystellardb >= 0.9 contains beeline installation to /usr/local/bin/beeline.

Requirements

Install using

• pip install 'pystellardb[hive]' for the Hive interface.

PyHive works with

• For Hive: HiveServer2 daemon

Windows Kerberos Configuration

Windows Kerberos configuration can be a little bit tricky and may need a few instructions. First, you’ll need to install & configure Kerberos for Windows. Get it from http://web.mit.edu/kerberos/dist/

After installation, configure the environment variables. Make sure the position of your Kerberos variable is ahead of JDK variable, avoid using kinit command located in JDK path.

Find /etc/krb5.conf on your KDC, copy it into krb5.ini on Windows with some modifications. e.g.(krb5.conf on KDC):

[logging]
default = FILE:/var/log/krb5libs.log
kdc = FILE:/var/log/krb5kdc.log
admin_server = FILE:/var/log/kadmind.log

[libdefaults]
default_realm = DEFAULT
dns_lookup_realm = false
dns_lookup_kdc = false
ticket_lifetime = 24h
renew_lifetime = 7d
forwardable = true
allow_weak_crypto = true
udp_preference_limit = 32700
default_ccache_name = FILE:/tmp/krb5cc_%{uid}

[realms]
DEFAULT = {
kdc = host1:1088
kdc = host2:1088
}

Modify it, delete [logging] and default_ccache_name in [libdefaults]:

[libdefaults]
default_realm = DEFAULT
dns_lookup_realm = false
dns_lookup_kdc = false
ticket_lifetime = 24h
renew_lifetime = 7d
forwardable = true
allow_weak_crypto = true
udp_preference_limit = 32700

[realms]
DEFAULT = {
kdc = host1:1088
kdc = host2:1088
}

Above is your krb5.ini for Kerberos on Windows. Put it at 3 places:

C:ProgramDataMITKerberos5krb5.ini

C:Program FilesMITKerberoskrb5.ini

C:Windowskrb5.ini

Finally, configure hosts file at: C:/Windows/System32/drivers/etc/hosts Add ip mappings of host1, host2 in the previous example. e.g.

10.6.6.96     host1
10.6.6.97     host2

Now, you can try running kinit in your command line!

Testing

On his way