DH

With the help of @prkamath's [answer](https://github.com/jupyter/jupyter/issues/293#issuecomment-462032713) and settings shown in [Nginx + notebook issue](https://github.com/jupyter/notebook/issues/625#issuecomment-362985148), we found this Nginx setting worked for us. Thanks! ``` map $http_upgrade $connection_upgrade { default upgrade;...

Thank you for your prompt reply, @rwedge ! I work with `fsspec==2021.8.1`. Any issue with this version? The link you mentioned greatly inspired me. As working with quite a large...

Thank you, @rwedge Basically, I'll DFS hundreds of times in the loop as demostrated below. Note the code worked as expected at first, but always crashed after some iterations. ```...

Your prompt and detailed reply is highly appreciated, and sorry for my late reply @rwedge * With your intuitive demo, I get dashboard work now. Hopefully it will do me...

Hi @rwedge , it seems there're still something wrong with what I did. ``` from dask.distributed import Client from dask.distributed import LocalCluster cluster = LocalCluster(n_workers=6) ft.dfs( ... dask_kwargs = {"cluster":...

Your prompt reply is highly appreciated, @thehomebrewnerd. Actually I put dfs and initiated cluster & client in a function, which was used in a loop. The code structure is shown...

Thanks @thehomebrewnerd! Yeah, I know. I also tried to pass cluster & client to ft_raw, as shown below, but I got **BrokenPipeError** (one time), and then **TimeoutError** (always). I doubt...

Thank you so much @thehomebrewnerd Yeah, maybe it was due to out of memory. After reducing n_workers from 6 to 2, and chunksize from 20000 to 10000, I got your...

@JanPalasek I got the same error even trying to convert an empty notebook with command: `jupyter nbconvert --to html --template pj --no-input notebooks/test.ipynb`