Extending Device Python Classes From Other ZenPacks

Recently I was struggling with trying to make two custom ZenPacks work together with device objects in the Zenoss database. My issue was that a main ZenPack (which would be installed first) would create a custom class that inherits from Device. This ZenPack would add a lot of properties and relationships with custom components. Then, a second optional ZenPack could be installed which would extend this same class with an additional property that would work along with a Python data source.

The zenpack.yaml file for the first and main ZenPack would look something like this:

# ZenPack1

    base: [zenpacklib.Device]
    label: Special Server

    base: [zenpacklib.Component]
    label: Special Component
      # ...

    base: [zenpacklib.Component]
    label: Amazing Component
      # ...

  - SpecialServer 1:MC SpecialComponent
  - SpecialServer 1:MC AmazingComponent

So what I wanted was to somehow be able to extend this SpecialServer class and add a new property to it, all of this done from the second ZenPack. Initially I tried doing this using ZenPackLib, doing something along the lines of:

# ZenPack2

        type: boolean
        default: false

Hoping that zenpacklib would somehow be smart enough to just append this additional attribute. The above however did not work.

ZenPack Monkey Patching

It turns out that the way to do this is to perform some monkey patching from the ZenPack. Monkey patching is the dynamic replacement of attributes at runtime.

When developing ZenPacks that perform monkey patching, there are some guidelines that can be followed for more organized code.

ZenPack Patches Directory

First we must create a directory called patches in the ZenPack's main directory (where zenpack.yaml resides). Inside the patches directory, we will place a Python init file that will take the responsibility of validating that everything that will be monkey patched can be correctly imported when the ZenPack is installed:


import logging
from importlib import import_module

log = logging.getLogger('zen.ZenPack2')

def optional_import(module_name, patch_module_name):
    except ImportError:
        except ImportError:
            log.exception('failed to apply %s patches', patch_module_name)

optional_import('ZenPacks.aalvarez.ZenPack1', 'ZenPack1')

For everything that we want to monkey patch, we will call the optional_import function with two arguments: The required import (in this case, the first ZenPack), and the name of the file that will contain the monkey patches (in this case ZenPack1). This will be created later and will reside in the same location as __init__.py.

After taking a look at official Zenoss ZenPacks that contain monkey patching, I discovered that monkey patches to the Zenoss Core source code should be placed inside a file called platform.py.

Since we used ZenPack1 as the name (you can use any name), we will created a new file called ZenPack1.py:


import logging
log = logging.getLogger('zen.ZenPack2')

from Products.Zuul.infos import ProxyProperty

from ZenPacks.aalvarez.ZenPack1.SpecialServer import SpecialServer, SpecialServerInfo

# Add new property
SpecialServer.my_new_property = False
SpecialServer._properties += (
    {'id': 'my_new_property', 'type': 'boolean', 'mode': 'w'},

# Make the property available through the API
SpcialServerInfo.my_new_property = ProxyProperty('my_new_property')

This file is directly monkeypatching the SpecialServer class from the ZenPack. To understand how these classes are constructed, you can take a look at the main parent class found at $ZENHOME/Products/ZenModel/Device.py.

Apart from the SpecialServer class, we are also importing its Info class to also assign the new property. This will make the property available through the API. Meaning that you can use the Zenoss JSON API to query the device information. This new property will be included in the JSON response results.

Info classes define a mapping between object attributes and interface classes for display in the GUI.

In ZenPacks, Info classes are located in info.py file. However when using ZenPackLib this is no longer necessary since ZenPackLib takes care of generating the necessary Info code from the YAML file.

The info.py file abstracts object attribute information saved in the Zope Object Database (ZODB), that will be displayed to the user.

Lastly we need to add all this monkey patching logic to the installation process. This is done at the end of the ZenPack's main __init__.py file:

# YAML loading here ...

# ...

# Patch last to avoid import recursion problems
from . import patches

And now our monkey patching is completed. When the ZenPack is installed, the new property will be added to ZenPack1's SpecialServer class.


  1. Monkey Patching in ZenPacks
  2. ZenPack Developer's Guide v1.0.1 - Jane Curry, page 176

Comments 32   Comments

comments powered by Disqus