Florian Jetter

I checked the behavior again. If `silence_logs=True` we're still setting all the loggers to silent but with `silent_logs=False` it is respecting the config set in the dask.config ctx manager. I...

Apologies for the slow reply. IIUC this deadlock can happen in pretty much every situation since our profiling thread is using `sys._current_frames()`. Do you know how easy it is to...

> I also do not immediately see where the This is a little hidden but it is happening while the `Future` object is instantiated (see [Future.bind_late](https://github.com/dask/distributed/blob/854a280dd8b6f485a0f682d4990b9f8c15012149/distributed/client.py#L241)) --- This race condition...

another mitigation that's already been mentioned is to ensure that the name of every scattered future is unique. Providing `hash=False` to `Client.scatter` will generate a unique key for every submission...

Can you provide an example of how you create your data? Typically the only thing you'll need to do for a result to be retained by the worker and for...

> The vaex dataframes were showing up as unmanaged memory and had poor performance when calling gather because of object introspection, so now we serialize the vaex dataframes to arrow...

This reminds me slightly of https://github.com/dask/distributed/issues/6792 which, to some part, also discusses how to manage hooks to cluster instances or how to simplify UX around this. The path-to-spec bit is...

Thanks for merging #39! I think we're ready for a release. Would you take care of this please @jrbourbeau ?

Apart from the code itself, I think a big-ish issue with the current setup is that I suspect our default deployment configuration is not ideal for the machine the benchmarks...

I wanted to follow up with some improvements but we can of course all do in one go. > To get the PR merged and the results updated quickly I...