Brett Naul

@Fleid @mikealfare this has been out there for a while now, I think it's basically impossible at this point that they would drastically change the functionality, as they have [written...

It appears that this is still an issue, in my case I'm seeing 502s not 504 but they also manifest as the same JSONDecodeError. Even just doing `request.raise_for_status()` before returning...

Thanks @jarredSafegraph, I would still say though that the error handling could be better for such cases (unless you think this will be the last ever 50* error a user...

Aggregating the various bokeh issues here.

See also https://github.com/cesium-ml/cesium/issues/92.

@jacobtomlinson I can definitely confirm that those workers are gone, the reason I started digging into this in the first place was that my `highmem` worker group kept getting 💥'd....

For now I am working around this with ``` class WorkerNameFilterPlugin(SchedulerPlugin): """Workaround for https://github.com/dask/dask-kubernetes/issues/914.""" def __init__(self): self.name = "worker-name-filter" self.idempotent = True def valid_workers_downscaling(self, scheduler, workers): return [worker for worker...

I tried registering a custom tokenizer function with dask and that seemed to speed things up by about 3x (not claiming this is necessarily a legit implementation, just a first...