Datera Python SDK

Introduction

This is Python SDK version 0.1 for the Datera Fabric Services API. Download and use of this package implicitly accepts the terms in COPYING

Users of this package are assumed to have familiarity with the Datera API. Details around the API itself are not necessarily covered through this SDK.

Installation

To install:

    git clone https://github.com/Datera/python-sdk.git
    cd python-sdk
    python setup.py install

Managed Objects

Datera provides an application-driven storage management model, whose goal is to closely align storage with a corresponding application's requirements.

The main storage objects are defined and differentiated as follows:

Application Instance (AppInstance)

-	Corresponds to an application, service, etc.
-	Contains Zero or more Storage Instances

Storage Instance

-	Corresponds to one set of storage requirements for a given AppInstance
-	ACL Policies, including IQN Initiators
-	Target IQN
-	Contains Zero or more Volumes

Volumes

-	Corresponds to a single allocated storage object
-	Size (default unit is GB)
-	Replication Factor
-	Performance Policies (QoS for Bandwidth and IOPS)
-	Protection Policies (Snapshot scheduling)

Another way of viewing the managed object hierarchy is as follows:

app_instances:
	- storage_instances:                 (1 or more per app_instance)
		+ acl_policy                     (1 or more host initiators )
		+ iqn                            (target IQN)
		+ ips                            (target IPs)
		+ volumes:                       (1 or more per storage_instance)
			* name
			* size
			* replication
			* performance_policy         (i.e. QoS)
			* protection_policy          (i.e. Snapshot schedules)

Endpoints

HTTP operations on URL endpoints is the only way to interact with the set of managed objects. URL's have the format:

      http://192.168.42.13:7717/v2/<object_class>/[<instance>]/...

where 7717 is the port used to access the API, and "v2" corresponds to an API version control.

Briefly, the REST API supports 4 operations/methods create (POST), modify (PUT), list (GET), delete (DELETE). Any input payload is in JSON format; any return payload is in JSON format. Login session keys are required within the "header" of any HTTP request. Sessions keys have a 15 minute lifetime.

For a full reference documentation of the REST API, please review the Datera REST API Guide.

This Python SDK serves as a wrapper around the raw HTTP layer.

Using this SDK

The Datera module is named dfs_sdk, and the main entry point is called DateraApi. Obtaining an object handle can be done as follows:

		from dfs_sdk import DateraApi
		[...]
        api = DateraApi(username=user, password=password, hostname=ipaddr)

Common Objects, Examples and Use Cases

Please see the utils directory for programming examples that cover the following:

Common methods for all objects include create(), set(), delete(), list()

• To create an app_instance with name FOO:

        ai = api.app_instances.create(name="FOO")

• Looping through objects can be done via list():

        for ai in api.app_instances.list():
		    print "AppInstance: ", ai

• To set a given app_instance into an offline state:

        ai.set(admin_state="offline")

• To delete a given app_instance:

        ai.delete()

'dhutil' : Datera Host Utility

The 'dhutil' host-utility is provided along with this SDK. 'dhutil' can be used as both a reference example for using the SDK, as well as providing some common host-side utility. For example, a given storage/application lifecycle might looks like this:

• Create 5 app_instances named 'mongodev', each with a single 10G volume, and to perform the host-side iscsi scan and login:

          dhutil --basename mongodev --count 5 --size 10

• View the multipath mapping to the host:

         dhutil --mpmap

• Create xfs filesystems for the 'mongodb' volumes, mount them at '/mnt' and change the permissions to 'mongodb:mongodb':

          dhutil --basename mongodev --mkfs --dirprefix /mnt --chown mongodb:mongodb

• Completely teardown (unmount, remove directory, iscsi logout, delete app_instances):

          dhutil --basename mongodev --cleanall

Note that steps 1 and 3 could be combined as follows:

       dhutil --basename mongodev --count 5 --size 10 --mkfs --dirprefix /mnt --chown mongodb:mongodb

Or a corresponding "app_template" could be used, if available:

       dhutil --basename mongodev --count 5 --template mongodb ...

Caveats

• dhutil presumes a 'singleton' model, whereby an app_instance is created with a single storage_instance with a single volume. Extending the functionality is left as an exercise for the reader and is strongly encouraged!

Reporting Problems

For problems and feedback, please email "[email protected]"