Update GRETA sync docs / process to show how to get 'default' packages there

[jeanconn]

Do we want to put all of the core packages we use in the icxc ska3-conda area or put these in a separate local repo that we'll sync?

One GRETA issue is that it looks like "conda index" comes from conda-build which isn't installed from Miniconda. This is an issue if you want to put files from two "source" directories into one repo on GRETA as if you can't reindex the new combined repo in the bare startup environment you can't update conda index easily to actually install core modules with dependencies.

[jeanconn]

I like everything to be orderly, so my preference would be to make a new file repo that has what we want from 'defaults' for GRETA, sync that with one more rsync command, and add it as a replacement for -defaults in the .condarc.

I don't have a good justification for a separate channel though. We could probably just as easily copy everything into ska3-conda/linux-64 and reindex.

[taldcroft]

👍 on the separate and orderly idea.

[jeanconn]

OK. I think the reasonable process would then be to install ska3-core and ska3_builder in a conda root environment (giving us all the core packages and conda-build and such at the right version), rsync the packages from PREFIX/pkgs/ to the new repo on HEAD, and index.

This might be a reasonable use for the builder environment itself.

[taldcroft]

👍 (I assume you don't get notification of me +1'ing your last comment).

[jeanconn]

OK. I needed a brand new conda environment to not have "pkgs" contaminated with other ska packages, so I made a /proj/sot/ska3/core and set the directions to do that. I deleted /proj/sot/ska/skare3_build.

We have a few too many envs hanging around anyway; will probably want to update the maintainers docs to talk about using a local conda install for speed as well.

[taldcroft]

talk about using a local conda install for speed as well.

👍 👍

I use a local /export dir on kadi for generic Python, but the idea of using it for Ska analysis is enticing. Another win for conda Skare3!

[jeanconn]

I was thinking about speed for building/installing (which I think does lots of small tasks that may have lock/unlock processes and a local file registry and such for the task), but I suppose it should be faster to work locally for analysis too.

[taldcroft]

Import times are dramatically faster.