docker pull getting very slow as our hpcnode image gets big

[zackramjan]

I think dockerhub has us throttled down to a few MB/s, so a pull on a new node deployment is taking like an hour in ansible.

we should probably set up a docker caching registry so that we dont need to hit docker from all nodes all that time.

it makes sense as to why they would throttle us, since we have a cronjob that looks like this:

30 20 * * * /usr/lpp/mmfs/bin/mmdsh -N all docker pull vanandelinstitute/hpc

so potentially nearly 70 pulls of a large image at once, ie hundreds of GB.

this is what we need to do most likely: https://docs.docker.com/docker-hub/mirror/