Hampus Rosvall
09/29/2022, 3:19 PMlp.py
but it is maybe the cleanest solution to this problem?
(.venv) ☁ flyte [main] tree .
.
├── <http://in_container.mk|in_container.mk>
├── launchplans
│ └── lp.py
└── workflows
├── wf.py
Haytham Abuelfutuh
09/29/2022, 7:04 PMHampus Rosvall
09/29/2022, 8:16 PMlp.py
then is
# imports
lp_definition_use_case_1
...
lp_definition_use_case_N
If I change launchplan_i, then I would need to re-deploy all launch plans. But maybe that’s fine?lp.py
and want to register only the launchplans, I should not do so with e.g., pyflyte register
since it is aimed for tasks and workflows as it requires an image?Samhita Alla
09/30/2022, 7:27 AMpyflyte register
for lp.py
as well: pyflyte register lp.py
, without sending an --image
argument if you’re okay with using the latest flytekit image.If I change launchplan_i, then I would need to re-deploy all launch plans. But maybe that’s fine?I believe that’s the right path for now. @Eduardo Apolinario (eapolinario) @Yee, any ideas?
Hampus Rosvall
09/30/2022, 7:37 AM--image
flag. What does it entail to deploy launch plans using the latest flytekit image then?Samhita Alla
09/30/2022, 9:20 AM--image
flag is required cause a launch plan executes a workflow of specific version which might have already been associated with an image (in your case).Hampus Rosvall
09/30/2022, 3:38 PMSamhita Alla
10/03/2022, 4:35 AMHampus Rosvall
10/03/2022, 1:47 PMpyflyte register flyte/launchplans/lp.py
- is there a way to only deploy a launch plan that runs the latest version of a workflow without deploying a new workflow version? What is best practice in terms of working with launch plans?
from flytekit import LaunchPlan
from ..workflows import wf
LaunchPlan.get_or_create(
workflow=<http://wf.wf|wf.wf>,
name="flytesnacks_lp",
default_inputs={},
)
launch_plan.yaml
Samhita Alla
10/05/2022, 5:59 AMHampus Rosvall
10/05/2022, 6:19 AMSamhita Alla
10/06/2022, 6:20 AMLaunchPlan.get_or_create
for you to associate your launch plan with a particular workflow cause we aren’t giving an already registered workflow as an input to it. Does this make sense?