In a python script, when updating an object being stored in runtime properties, for example:

intial script:

ctx.instance.runtime_properties['a_list'] = []
ctx.instance.runtime_properties['a_dict'] = {}

then, in execution of another script:

ctx.instance.runtime_properties['a_list'].append(1)
ctx.instance.runtime_properties['a_dict'].update({
    'key1': 'value1'
})

node instance will not be updated, and the changes made will not be saved.

If we did additionally any assignment to runtime properties:

ctx.instance.runtime_properties['a_list'].append(1)
ctx.instance.runtime_properties['a_dict'].update({
    'key1': 'value1'
})
ctx.instance.runtime_properties['a_var'] = 'some_value'

the internal "dirty" flag of runtime properties object will be set and node instance will have runtime properties updated according to changes made.

I can see the whole mechanism of detecting changes to runtime properties, being object of DirtyTrackingDict is based on setting the .dirty flag, but only in case of direct assignments or update method call.
If we do the same on objects which are nested in runtime properties, the mechanism does not work and probably it will be hard to make it working in current design.
A potential solution to it could be:

• before executing the script, make a deepcopy of runtime properties
• execute the script
• compare runtime properties with the copy done intially

initial_runtime_properties = copy.deepcopy(ctx.instance.runtime_properties)
run_script(ctx, ...)
if ctx.instance.runtime_properties != initial_runtime_properties:
    ctx.instance.runtime_properties._set_changed()  # what sets .dirty=True

I got bugs:
task failed 'cloudify_cloudinit.tasks.update' -> 'module' object has no attribute 'zmqlogginghandler' [retry 35/60].
Please help me to fix that.
Thanks!

Hi, is there any possibility to implement "kill_cancel" in cloudify-rest-client? Or how should I handle hung/stuck force_cancel in my test automation?

Regards,
Karol

Pika=="0.11.2" dependency makes it impossible to use with Python version >= 3.7. That's because pika in version used currently uses async as a properites name which is currently keyword in Python. The solution can be to upgrade pika dependency to 1.1.0 version.