Ajay
09/03/2024, 8:04 AMMarvin
09/03/2024, 8:04 AMAjay
09/03/2024, 8:05 AMMarvin
09/03/2024, 8:07 AMDaskTaskRunner
accepts an address
parameter to connect to an existing Dask scheduler and other parameters like cluster_class
, cluster_kwargs
, etc., to customize your Dask cluster. However, it does not directly support dynamically changing the client address mid-flow.
- Prefect Task Runners
- Usage Guide
2. GitHub Issues: There isn't direct support for switching the Dask client address on-the-fly within a single flow. However, you can manage separate Dask clusters and configure your flows to connect to them as needed. The relevant GitHub issues also don't provide a ready-made solution but indicate potential workarounds or configurations.
- HPC Machines and Job Schedulers Issue
- Coiled Cluster Issue
In conclusion:
- You cannot directly change the Dask client address mid-flow.
- You might need to architect your flow to segregate tasks according to the cluster requirements or handle custom connections programmatically.
Do these glimmers of hope align with your ambitious endeavor? Or does your chaotic dream need more grounding in reality?