Proper shutdown for dask.distributed
- 主要语言
- Jupyter Notebook
- 星标
- 2.6k
- 派生
- 1k
- PR 合并指标
- 30 天内没有已合并 PR
描述
When using the [`stop_dask` method]( http://ipyparallel.readthedocs.io/en/latest/api/ipyparallel.html#ipyparallel.Client.stop_dask ) of the `ipyparallel.Client`, the following warning message pops up.
```
distributed.client - WARNING - Client report stream closed to scheduler
```
However, if I run [`shutdown`]( http://distributed.readthedocs.io/en/latest/api.html#distributed.client.Client.shutdown ) on the Dask client provided by [`become_dask`]( http://ipyparallel.readthedocs.io/en/latest/api/ipyparallel.html#ipyparallel.Client.become_dask ), and then run `stop_dask`, the warning goes away.
Looking at the [code in ipyparallel]( https://github.com/ipython/ipyparallel/blob/6.0.2/ipyparallel/client/client.py#L1334 ) it is not obvious how it would handle calling `shutdown` as it does not know the Dask client in `stop_dask`. However it does seem `shutdown` needs to be called on the `executor`. How exactly should shutdown be done with `dask.distributed`?
贡献指南
评估
这个 Issue 还没有评估数据。