Guido Imperiale
Guido Imperiale
...ok, I found an issue that may scupper the whole design. In the first part of `Worker.execute`, there are many accesses to `ts` which require the task to not have...
I fixed the race condition I described in my previous post. The fix is... *fancy*, and testing it requires some finesse. At first sight, this detracts from the key point...
We've got another issue automatically fixed by this PR: #6877.
> #6869 is an example why I am concerned about letting multiple coroutines for the same task running, i.e. gather_dep and execute. I can't understand how this PR would make...
I've removed the `done` attribute :metal:
I've found a roadblock that I didn't consider before. When you transition from executing to released, all dependencies are immediately released as well (unless there are other dependants). However, they...
This community issue is likely impacted by this PR: https://dask.discourse.group/t/slow-down-transfer-between-workers/1012/2
I'm leaving the review of the transition machinery and tests to @fjetter. Besides rounding, my perplexity entirely revolves around the `is_rootish` method. 1. Why do we need it to begin...
I agree we should aim to get this in main quickly and then further iterate. I can see two avenues: 1. get it in as-is (post minor tweaks), and then...
+1. I like the idea to use the OS if possible and only fall back on the nanny polling system if not available.