Jason Parraga

@machichima do you still want to land this? If so I can take a look and we can merge this.

More concretely, it looks like this function has some hardcoded `nvidia.com/gpu` which seems to be taking precedent over what we are configuring: https://github.com/flyteorg/flyte/blob/6a1efa58599874038ccbbed7df5ad0bcaf53c1a5/flytepropeller/pkg/controller/nodes/task/taskexec_context.go#L194-L210

Support for the concurrency limits came in flytekit 1.16.6: https://github.com/flyteorg/flytekit/releases/tag/v1.16.6 If what you're saying is true I don't see how the following unit tests could have passed: https://github.com/flyteorg/flytekit/pull/3267/files#diff-c4cfe855bcb35511ef7e3bfde95a23a24f49465fcf8adb81280213e3e4ddfdd3R7

I can verify that I was seeing the same error until I updated `flyteidl` to `v1.16.1` so I have to believe this is a user error. v1.15.3 ``` /Users/jasonparraga/code/flytekit/.venv/bin/python /Applications/PyCharm.app/Contents/plugins/python-ce/helpers/pycharm/_jb_pytest_runner.py...

@mhotan can you fix the conflicts?

I'm not sure if this is the correct approach. For regular Python tasks the timeout is passed through the task metadata and queueing time is counted towards the overall task...

We just upgraded from flytekit v1.14.6 to v1.15.4 and all tasks using pod_template with_overrides seems to be broken. We are seeing bad task specification errors in propeller because the pod...

This feature also doesn't work when we eliminate the `pod_template` from the task annotation and only set it using `with_overrides`.

> Hi @Sovietaced, May I ask what your flyte version is? V1.15.3 on the server

> I created a workflow using your pseudocode with Flyte v1.15.3 and Flytekit v1.15.4, but I still can’t reproduce your error. > My example code is below. Is there anything...