C.J. Collier

I have tried using the file:// prefix, I thought! I'll review download.py and see what I can come up with.

here's the code where I set the channel mirror: https://github.com/LLC-Technologies-Collier/custom-images/blob/metadata-inclusion-20241029/examples/secure-boot/rapids.sh#L710 ``` for channel in 'rapidsai' 'nvidia' 'pkgs/main' 'pkgs/r' 'conda-forge' ; do "${CONDA}" config --set \ "custom_channels.${channel}" "file://${rapids_mirror_mountpoint}/conda.anaconda.org/" done ``` Is...

Here's the script I use to populate `file://${rapids_mirror_mountpoint}/conda.anaconda.org/`: https://github.com/cjac/dataproc-repro/blob/conda-mirror-20241031/lib/mirror/sync-conda.pl

Hmm... that code looks like it prefers `file:` to `file://`... oh, you said *not* using the file:// prefix! I'll give that a go.

conda won't accept `file:/srv/mirror/...` or `/srv/mirror/...` as config values instead when running `conda config --set custom_channels.conda-forge file:///srv/mirror/conda.anaconda.org` I tried changing the condition to use local if file:// in self.url instead....

> it simply ignores it Yes, that's nearly my experience as well. It does seem to parse the url for something that looks like a hostname and uses that. In...

it looks like appending /srv/mirror/conda.anaconda.org/*/{linux-64,noarch} to the pkgs_dirs lists prevents the copy to the temp directory! It also spews tons of errors when performing the install, but we can redirect...

strange. I've built from scratch on 2.2 and 2.3 very recently. If you can increase the size of your VM to 32 cores, use a single instance cluster, and create...

@zafercavdar - have you been able to confirm that the script from #1320 resolves your issue? Here is a direct link to that file: https://github.com/GoogleCloudDataproc/initialization-actions/raw/6d00e017e6c46a91d646cb3ea32c78925a3f7474/gpu/install_gpu_driver.sh

/gcbrun