Task is pending with reason "user env retrieval failed required held"

Get Help
1

Hello!
I’m using OpenOnDemand 4.0.8 and Slurm 25.11.3

When running jobs via job composer, the job is queued and assigned a pending status with the reason “user env retrieval failed required held.”
This is because the sssd service configuration file on the compute node contains the following setting:

access_provider = ldap
ldap_access_filter =(memberOf=cn=admins,ou=groups,dc=hpc,dc=frccsc,dc=ru)

If I comment out these lines, the job starts and runs.
However, there’s something confusing. When running a job from the Slurm Login host, it always runs, and there’s no need to comment out these lines.

How should I configure OpenOnDemand to allow the job to run with the specified sssd service settings?

I’m attaching the OpenOnDemand settings:

cluster.yml

---
v2:
  metadata:
    title: "FRCCSC HPC"
  login:
    host: "login.hpc.frccsc.ru"
  job:
    adapter: "slurm"
    bin: "/usr/bin/"
    conf: "/etc/slurm/slurm.conf"
    copy_environment: true

defaul.sh (job file created via job composer)

#!/bin/bash

#SBATCH --partition=main
#SBATCH -w g5500-1
#SBATCH --ntasks=1
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=1
#SBATCH --cpus-per-task=2
#SBATCH -o slurm.out

######## It is NOT recommended to change settings in this section ########
export SLURM_EXPORT_ENV=ALL
##########################################################################

## Run the program
date && hostname

I am ready to provide additional information if necessary.

Thank you!
Best regards!

2

This is not an Open OnDemand issue as we don’t do anything with auth for slurm, we just issue sbatch commands on the login node.

But, I’m pretty sure I can see the problem:

ldap_access_filter =(memberOf=cn=admins

You’ll have to have allow other users besides admins to login, that’s likely why you can ssh and it works (guessing you are an admin here) but OOD can’t, however I’m not a slurm expert nor an sssd expert. Just a guess looking at the filter you set.