@ancient-wolf-19325 what does this error mean

Remote output size exceeds max

The error

Remote output size exceeds max

means that a Flyte task tried to write an output file (such as outputs.pb) that is larger than the maximum allowed size configured for remote outputs. When this happens, FlytePropeller fails the task with an

OutputSizeExceeded

error, and the workflow may get stuck or fail if the error is not properly propagated. The default max output size is often set to 10MB, but this can be configured in FlytePropeller settings. See example and discussion in Flyte Issue #381 and the relevant handler code. Would you like more detail on how to resolve or configure this limit? Sources: - Flyte Issue #381 - handler.go (FlytePropeller) 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.

@ancient-wolf-19325 what is the output usually ? is log an output ?

In Flyte, the output of a task or workflow is typically the data explicitly returned by the Python function, such as files, dataframes, or primitive types—these are strongly typed and passed between tasks as outputs. Logs are not considered outputs; they are separate artifacts accessible via the Flyte UI or cloud provider log links, but are not passed as workflow outputs or used as downstream task inputs. See Task input and output and . Would you like more detail on how outputs and logs are handled or accessed? Sources: - Task input and output - Figure Image 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.

@ancient-wolf-19325 do you have any example on how to set the max output size ?

Yes. To set the max output size for remote outputs in Flyte, configure the

storage.limits.maxDownloadMBs

parameter in your Flyte deployment configuration. For example, in a Helm values file, you can add:

configuration:
  inline:
    storage:
      limits:
        maxDownloadMBs: 100

This sets the maximum allowed output size to 100MB. See Flyte Issue #4951 for details. Would you like more detail on where to place this configuration or how it interacts with other settings? Sources: - Flyte Issue #4951 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.

what does flytepropeller do @ancient-wolf-19325?

FlytePropeller is the Kubernetes controller in Flyte responsible for scheduling and tracking the execution of workflows, which are represented as Directed Acyclic Graphs (DAGs) of tasks and nodes. It manages workflow state transitions, launches Kubernetes pods for tasks, and uses plugins to execute different task types, ensuring workflows progress from start to completion. See FlytePropeller architecture. Would you like more detail on its internal components or how it interacts with other Flyte services? Sources: - FlytePropeller architecture 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.