-
Notifications
You must be signed in to change notification settings - Fork 117
Description
I have configured Slurm with Pyxis extension and Enroot containers. In /etc/slurm/plugstack.conf/pyxis.conf, I set the following configuration:
required /usr/local/lib/slurm/spank_pyxis.so runtime_path=/tmp/enroot_data
The directory /tmp/enroot_data exists and is readable/writable, however, when importing Docker images, it fails with the error:
[2025-11-06T20:36:19.163] [19.0] error: pyxis: child 70956 failed with error code: 1
[2025-11-06T20:36:19.163] [19.0] error: pyxis: failed to import docker image
[2025-11-06T20:36:19.163] [19.0] error: pyxis: printing enroot log file:
[2025-11-06T20:36:19.163] [19.0] pyxis: could not remove squashfs /tmp/enroot_data/1000/19.0.squashfs: No such file or directory
[2025-11-06T20:36:19.163] [19.0] error: pyxis: couldn't start container
[2025-11-06T20:36:19.163] [19.0] error: spank: required plugin spank_pyxis.so: task_init() failed with rc=-1
[2025-11-06T20:36:19.163] [19.0] error: Failed to invoke spank plugin stack
Steps to Reproduce:
-
Configure Pyxis with runtime_path=/tmp/enroot_data in plugstack.conf
-
Submit a Slurm job with container support:
bash
#SBATCH --container-image=ubuntu
- Job fails during container image import phase
What I've Tried:
-
Confirmed /tmp/enroot_data directory exists with proper read/write permissions
-
Verified Slurm user can access the directory
-
The directory structure shows /tmp/enroot_data/1000/ but the specific squashfs file mentioned in the error doesn't exist.
Could you please tell me what the problem might be? Thanks in advance.