Bootstrap engine via python without the default site config

Pipeline Integrations
, , ,
1

Hey everyone,

I am currently working on running the shotgrid publisher on the farm.

It involves starting the software, in this case maya, and bootstrapping the tk-maya engine with the right context.

Currently I have the issue that shotgrid complains that it does not find the engine instance in the env/site.yml file when I try to start it.

I then went ahead and added a tk-maya entry in the site.yml file under engines.

includes:
- ./includes/frameworks.yml
- ./includes/settings/tk-desktop.yml
- ./includes/settings/tk-desktop2.yml
- ./includes/settings/tk-maya.yml

################################################################################
# configuration for all engines to load in a project context
engines:
  tk-desktop: '@settings.tk-desktop.site'
  tk-desktop2: '@settings.tk-desktop2.site'
  tk-maya: '@settings.tk-maya.site'

I also added the “site” section in the env/includes/settings/tk-maya.yml file.

settings.tk-maya.site:
  apps:
    tk-multi-publish2: "@settings.tk-multi-publish2.maya.shot_step"
  location: "@engines.tk-maya.location"

In the apps section I added the tk-multi-publish2 entry and set it up to use the shot_step settings as I am running a shot publish on the farm.

Now this feels very dirty to me because I would like to bootstrap the tk-maya engine in my python script with the already defined settings.tk-maya.shot_step settings.

I don’t want to use the default site config that I have to then adjust in the yml files to work with my specific shot case.

But I didn’t find a way how to do that. When bootstrapping it seems to always use the site config as default?

Can I somehow control that? This is how I currently do the bootstrap:

# Set up logging.
sgtk.LogManager().initialize_base_file_handler("tk-batch-rendering")

user = sgtk.authentication.ShotgunAuthenticator().create_script_user(
    os.environ["SHOTGUN_SCRIPT_NAME"],
    os.environ["SHOTGUN_SCRIPT_KEY"],
    os.environ["SHOTGUN_HOST"]
)

# Set the right pipeline configuration.
tk_manager = sgtk.bootstrap.ToolkitManager(user)
tk_manager.plugin_id = "basic.batch"
tk_manager.pipeline_configuration = configuration_id

# Sync the path cache before the engine starts. The engine validates the templates
# when it starts so path cache has to be already synced.
tk_manager.pre_engine_start_callback = lambda ctx: ctx.sgtk.synchronize_filesystem_structure()

# Start the engine, we're done!
tk_manager.bootstrap_engine(engine_instance_name, context)

Would love to get some tips!

2

Remove the site sections you added.
You could try defining the context your bootstrapping in to. This could be supplied to the deadline job as key/info fields. You will need the entity type and id.
If you don’t supply a context, sg will only bootstrap to the site context.

1 Like
3

Hey @Patrick

Thanks for the quick reply.

Yeah the weird thing is I am bootstrapping with a pre defined context with:

tk_manager.bootstrap_engine(engine_instance_name, context)

And in my case the context is dictionary of a task entity.
{'id': 39148, 'name': 'animation', 'type': 'Task'}

This should be sufficient right?

4

hmm it should be yes.
Did you make any breakthroughs?