Fix for AWS ECR issue not released?

[gallanik]

I have successfully integrated (proof-of-concept) enroot with AWS ECR after installing enroot from source (referring to #159).
Just wanted to check when can this is officially released? We have a production deployment of DGX SuperPOD and would like to install a stable release rather than from source. There hasn't been a release since more than an year now (Feb 8, 2023).

[RHudsonH]

I'm struggling with this as well. Any news? Looks like it's been more than a year since the fix was in main.

[zyndagj]

Assuming the main branch passes all legacy tests, it would be great if a release could be tagged with this fix.

[3XX0]

The fix is not correct, see discussion #159 (comment)
We need a proper fix and make sure it is validated

[3XX0]

Just committed 6425a53 which hopefully does the trick

[gallanik]

I will test this, but the earlier fix also worked without any issues. Any plans on validating and releasing this soon?

[astrophys]

I tested the commit 6425a53 as a hot fix on a BCM-10 cluster, and still got the error :

user@head01:~$ srun --container-image=docker://123456789.dkr.ecr.us-west-2.amazonaws.com#hello-world --nodelist=node02 --pty bash
pyxis: importing docker image: docker://123456789.dkr.ecr.us-west-2.amazonaws.com#hello-world
slurmstepd: error: pyxis: child 737010 failed with error code: 1
slurmstepd: error: pyxis: failed to import docker image
slurmstepd: error: pyxis: printing enroot log file:
slurmstepd: error: pyxis:     [INFO] Querying registry for permission grant
slurmstepd: error: pyxis:     [INFO] Authenticating with user: AWS
slurmstepd: error: pyxis:     [INFO] Using credentials from file: /home/user/.config/enroot/.credentials
slurmstepd: error: pyxis:     [INFO] Fetching image manifest list
slurmstepd: error: pyxis:     [ERROR] Could not process JSON input
slurmstepd: error: pyxis: couldn't start container
slurmstepd: error: spank: required plugin spank_pyxis.so: task_init() failed with rc=-1
slurmstepd: error: Failed to invoke spank plugin stack
srun: error: node02: task 0: Exited with exit code 1

Based on conversations I've had with others, it seems like ECR has some non-standard registry configuration.