Dave Trudgian comments

Results 233 comments of


                                            Dave Trudgian

Try unauthenticated OCI registry interactions if expired/invalid credentials are set

Currently we have the env vars `DOCKER_LOGIN` `DOCKER_USERNAME` `DOCKER_PASSWORD`, so I'd suggest probably something fairly explanatory e.g `DOCKER_TRY_NOAUTH` or `DOCKER_NOAUTH_FIRST` ... and probably exposed as a flag too `--docker-try-noauth`? Open...

Support for named volumes

In my view, this one is a bit difficult. It's definitely something that would be nice to have, but the devil is in the details. For example, in Docker, a...

Support for named volumes

If an explicit `volume create` being required in order to use a volume is an acceptable trade-off, and anonymous volumes are not implemented, then this is relatively straightforward. So long...

singularity run --nv doesn't give access to A100 on Ubuntu 20.04 host w/ CUDA 11.4

Thanks for the information. Noting here that you reported in Slack that it also works as expected when using `--nvccli` functionality - just not `--nv` alone.... which should narrow it...

singularity run --nv doesn't give access to A100 on Ubuntu 20.04 host w/ CUDA 11.4

Thanks for the additional detail. I am now somewhat confused why things aren't working... since all the libraries are being found and bound as expected. I am downloading the large...

singularity run --nv doesn't give access to A100 on Ubuntu 20.04 host w/ CUDA 11.4

Pulled the image locally onto my machine with space, and it appears to work.... ``` $ singularity run --nv e4s-gpu-x86_64.sif Singularity> nvidia-smi Mon Jan 24 10:12:12 2022 +-----------------------------------------------------------------------------+ | NVIDIA-SMI...

singularity run --nv doesn't give access to A100 on Ubuntu 20.04 host w/ CUDA 11.4

Thanks for the further details. I'll try it out with exactly that driver package when I get a chance, to see if I can replicate. Since it hasn't been reported...

singularity run --nv doesn't give access to A100 on Ubuntu 20.04 host w/ CUDA 11.4

An update on this. I couldn't replicate with the version of the driver reported here, on CE 3.9.5 before I just released it. Still not sure of the cause.

singularity run --nv doesn't give access to A100 on Ubuntu 20.04 host w/ CUDA 11.4

Thanks for the note @michaelkarlcoleman I'm wondering now whether the original issue is related to https://stackoverflow.com/questions/69115621/cuda-error-not-initialized-on-a100-after-server-reset In the first post here where the issue was reported, the host `nvidia-smi` output...

Can't pull signed images from registry where /etc/containers/policy.json enforces signature checks

Thanks for the information. It looks like the behavior of the `containers/image` dependency that we use to retrieve and manage OCI layers prior to conversion into Singularity's SIF formats may...