Chris Roat

Thanks for the info. The [gcloud command](https://github.com/seung-lab/igneous#executing-tasks-in-the-cloud) for creating the cluster in the README didn't use pre-emptibles. I will send an update for that. Is there anything else in that...

Are there any logs or monitoring that would point me to the task timing? The deployment.yaml has LEASE_SECONDS of 800.

The cluster is running OK by using with 2 core jobs, but I'd like to make everything a bit more optimal. I'm writing my original cloudvolume via dask, and then...

I'm using the default that comes with `create_transfer_tasks`. I don't see a setting there or in the sample yaml. Did I miss it? My bigger initial concern is the initial...

These are uint16 images. Its 5 channels because we are measuring florescence signal at different wavelengths. We actually sample many more (after different chemistry is done), but the acquisitions are...

We might be talking past each other. I have no timing info on the downsampling, which is done on an igneous cluster. My first is about the transfer task and...

Yeah, my quick-and-dirty change to turn of compression (assuming nothing else changed in the environment), means my initial outputs are 6s each, rather than 300s. Of course, it's 3x bigger!...

The igneous tasks do a 2048x2048x64 fetch (the default) and downsample with chunks 128x128x16 -- 4 mip levels. Just looking at how long between a message is first taken in...

Thanks for the extensive numbers.   On the creation of the original chunks from my dask pipeline, I'm making them 512x512x128.  This is the point where I found the gzip was...

Dask and Igneous do not interact. I was worried we kept talking past each other. There are two things here: 1. Dask writes out the original volumes (512x512x128). It's super-duper...