Riley Hun

08/26/2020, 1:19 AM
Hi everyone, Hope you're having a good week so far. Pardon my ignorance, but just a quick question - I am using a static dask cluster hosted on Google Kubernetes Engine on GCP and I have it set to auto-scale depending on CPU usage. Now, I am running a mapped task to unzip and extract some files from several zip folders, but I'm noticing that the number of Dask workers remain at 1. I guess my question is should I expect Prefect mapped tasks to spawn additional Dask workers to accomplish the parallelization? Checking the logs, it looks like it's extracting the files sequentially even though I am using the mapped task.

Marwan Sarieddine

08/26/2020, 3:46 AM
Hi Riley, a static cluster will only deploy a fixed number of workers - regardless if you have a node autoscaler enabled on your kubernetes cluster

Riley Hun

08/26/2020, 4:35 PM
Hi Marwan, I'm not just using a node auto-scaler for the kubernetes cluster. I have a scaler for the dask workers as well. Typically, they would adapt if the work load demanded it. Though that doesn't seem to be the case here for Prefect.