I've been trying to plot many figures using matplotlib+dask through the package xmovie. I've experimented with expressing memory limits and such is two ways and have gotten different results. For example if I use
client = Client(n_workers=18, memory_limit='0.5GB', processes=False)
parallel_plotting.compute()
I keep getting warnings like distributed.utils_perf - WARNING - full garbage collections took 90% CPU time recently (threshold: 10%) and even occasional errors.
However, if I pass the arguments to the compute function such as
parallel_plotting.compute(n_workers=18, memory_limit='0.5GB', processes=False)
everything seems to work fine.
Could someone please explain why these two ways of passing keywords are different? I have not been able to find the answer in the docs.