MuranoPL extension plug-ins¶
Murano plug-ins allow extending MuranoPL with new classes. Therefore, using such plug-ins applications with MuranoPL format, you access some additional functionality defined in a plug-in. For example, the Magnum plug-in, which allows murano to deploy applications such as Kubernetes using the capabilities of the Magnum client.
MuranoPL extension plug-ins can be used for the following purposes:
Providing interaction with external services.
For example, you want to interact with the OpenStack Image service to get information about images suitable for deployment. A plug-in may request image data from glance during deployment, performing any necessary checks.
Enabling connections between murano applications and external hardware
For example, you have an external load balancer located on a powerful hardware and you want your applications launched in OpenStack to use that load balancer. You can write a plug-in that interacts with the load balancer API. Once done, add new apps to the pool of your load balancer or make any other configurations from within your application definition.
Extending Core Library class functionality, which is responsible for creating networks, interaction with murano-agent, and others
For example, you want to create networks with special parameters for all of your applications. You can just copy the class that is responsible for network management from the Murano Core library, make the desired modification, and load the new class as a plug-in. Both classes will be available, and it is up to you to decide which way to create your networks.
Optimization of frequently used operations. Plug-in classes are written in Python, therefore, the opportunity for improvement is significant.
Murano provides a number of optimization opportunities depending on the improvement needs. For example, classes in the Murano Core Library can be rewritten in C and used from Python code to improve their performance in particular use cases.
MuranoPL package type plug-ins¶
The only package type natively supported by Murano is MuranoPL. However, it is possible to extend Murano with support for other formats of application definitions. TOSCA CSARs and HOT templates are the two examples of alternate ways to define applications.
Package types plug-ins are normal Python packages that can be distributed through PyPI and installed using pip or its alternatives. It is important that the plug-in be installed to the same Python instance that is used to run Murano API and Murano Engine. For multi-node Murano deployments, plug-ins need to be installed on each node.
To associate a plug-in with a particular package format, it needs to have a special record in [entry_points] section of setup.cfg file:
io.murano.plugins.packages = Name/Version = namespace:Class
[entry_points] io.murano.plugins.packages = Cloudify.TOSCA/1.0 = murano_cloudify_plugin.cloudify_tosca_package:CloudifyToscaPackage
This declaration maps particular pair of format-name/version to Python class that implements Package API interface for the package type. It is possible to specify several different format names or versions and map them to single or different Python classes. For example, it is possible to specify
[entry_points] io.murano.plugins.packages = Cloudify.TOSCA/1.0 = murano_cloudify_plugin.cloudify_tosca_package:CloudifyToscaPackage Cloudify.TOSCA/1.1 = murano_cloudify_plugin.cloudify_tosca_package:CloudifyToscaPackage Cloudify.TOSCA/2.0 = murano_cloudify_plugin.cloudify_tosca_package:CloudifyToscaPackage_v2
A single Python plug-in package may contain several Murano plug-ins including of different types. For example, it is possible to combine MuranoPL extension and package type plug-ins into a single package.
Tooling for package preparation¶
Some package formats may require additional tooling to prepare package ZIP archive of desired structure. In such cases it is expected that those tools will be provided by plug-in authors either as part of the same Python package (by exposing additional shell entry points) or as a separate package or distribution.
The only two exceptions to this rule are native MuranoPL packages and HOT packages that are built into Murano (there is no need to install additional plug-ins for them). Tooling for those two formats is a part of python-muranoclient.
Package API interface reference¶
Plug-ins expose API for the rest of Murano to interact with the package by implementing murano.packages.package.Package interface.
def __init__(self, format_name, runtime_version, source_directory, manifest):
format_name: name part of the format identifier (string)
runtime_version: version part of the format identifier (instance of semantic_version.Version)
source_directory: path to the directory where package content was extracted (string)
manifest: contents of the manifest file (string->string dictionary)
Note: implementations must call base class (Package) initializer passing the first three of these arguments.
Abstract properties that must be implemented by the plug-in:
Fully qualified name of the package. Must be unique within package scope of visibility (string)
Package version (not to confuse with format version!). An instance of semantic_version.Version
List (or tuple) of MuranoPL class names (FQNs) that package contains
Dictionary of requirements (dependencies on other packages) in a form of key-value mapping from required package FQN string to SemVer version range specifier (instance of semantic_version.Spec or string representation supported by Murano versioning scheme)
Package type: “Application” or “Library”
Human-readable name of the package as presented to the user (string)
Package description (string or None)
Package author (string or None)
Package supplier (string or None)
List or tags for the package (list of strings)
Package (application) logo file content (str or None)
Package (application) supplier logo file content (str or None)
YAML-encoded string containing application’s form definition (string or None)
Abstract methods that must be implemented by the plug-in:
def get_class(self, name):
Returns string containing MuranoPL code (YAML-encoded string) for the class whose fully qualified name is in “name” parameter (string)
def get_resource(self, name):
Returns path for resource file whose name is in “name” parameter (string)
Properties that can be overridden in the plug-in:
Canonical format name for the plug-in. Usually the same value that was passed to class initializer
Format version. Usually the same value that was passed to class initializer (semantic_version.Version)
Package file (.zip) content (str)
Usually, there is no need to manually implement all the methods and properties described. There is a murano.packages.package.PackageBase class that provides typical implementation of most of required properties by obtaining corresponding value from manifest file.
When inheriting from PackageBase class, plug-in remains responsible for implementation of:
This allows plug-in developers to concentrate on dynamic aspects of the package type plug-in while keeping all static aspects (descriptions, logos and so on) consistent across all package types (at least those who inherit from PackageBase).