Open
Description
What happened:
When creating a LocalCluster
object the comm is started on a random high port, even if there are no other clusters running.
What you expected to happen:
Should use port 8786
.
Minimal Complete Verifiable Example:
$ conda create -n dask-lc-test -c conda-forge -y python=3.8 ipython dask distributed
$ conda activate dask-lc-test
The dask-scheduler
command works as expeted.
$ dask-scheduler
distributed.scheduler - INFO - -----------------------------------------------
distributed.scheduler - INFO - -----------------------------------------------
distributed.scheduler - INFO - Clear task state
distributed.scheduler - INFO - Scheduler at: tcp://10.51.100.43:8786
distributed.scheduler - INFO - dashboard at: :8787
But LocalCluster
does not.
>>> from dask.distributed import LocalCluster
>>> cluster = LocalCluster()
>>> cluster.scheduler_address
'tcp://127.0.0.1:45083'
Anything else we need to know?:
- Dask version: 2020.12.0 (tried latest source too)
- Python version: 3.8
- Operating System: macOS and Ubuntu 18.04
- Install method (conda, pip, source): conda and source