Skip to content

Authentication issues

Omnia runs OpenLDAP in the OIM-hosted omnia_auth Podman container and configures SSSD on supported Slurm and login nodes. The LDAP search base is derived from SYSTEM_DOMAIN_NAME; it is not a fixed dc=omnia,dc=local value.

Establish a baseline

Check the server on the OIM:

Run on: OIM
source /etc/profile.d/omnia-env.sh
systemctl status omnia_auth --no-pager
podman ps -a --filter name=omnia_auth
podman logs --tail 100 omnia_auth
podman exec omnia_auth ldapsearch -x -H ldap://127.0.0.1 \
  -b '' -s base namingContexts

The final command mirrors Orchestrator's local container health check. It proves that the LDAP endpoint responds inside the container; verify the configured StartTLS or LDAPS path separately from a provisioned client.

Check a provisioned client:

Run on: affected Slurm or login node
systemctl status sssd --no-pager
authselect current
grep -E 'ldap_uri|ldap_search_base|ldap_id_use_start_tls|ldap_tls_cacert' \
  /etc/sssd/sssd.conf
getent passwd <ldap-user>

Do not check the host slapd service on the OIM. OpenLDAP is owned by the omnia_auth Quadlet service in this release.

omnia_auth is not running

Symptom

systemctl status omnia_auth is failed or inactive, the container is absent, or clients cannot contact the port selected by ldap_connection_type (TCP 389 for TLS or TCP 636 for SSL).

Resolution
  1. Inspect the service and container logs:

    Run on: OIM
    systemctl status omnia_auth --no-pager
    journalctl -u omnia_auth -b -n 200 --no-pager
    podman logs --tail 200 omnia_auth
    
  2. Confirm that the selected catalog enables OpenLDAP and that the active Orchestrator credential file contains the configured OpenLDAP username and password. Do not print the decrypted password into a terminal log.

  3. Rerun the supported preparation and validation flows:

    Run on: OIM host
    cd <OMNIA_SOURCE_PATH>/src/main
    ./omnia.sh --run orchestrator --tags prepare
    ./omnia.sh --run orchestrator --tags validate-deployment
    
    Run on: OIM host
    source /opt/omnia/activate-omnia.sh
    cd <OMNIA_SOURCE_PATH>/src/orchestrator/playbooks
    ansible-playbook orchestrator.yml --tags prepare
    ansible-playbook orchestrator.yml --tags validate-deployment
    

LDAP user is not found on a node

Symptom

id <ldap-user> or getent passwd <ldap-user> reports no entry even though the user exists in LDAP.

Cause

The user may be outside the configured search base, may lack POSIX attributes, or SSSD may have stale cached data or an incorrect generated configuration.

Resolution
  1. Read the actual search base from the node and query that base on the OIM. Replace placeholders with the values from the active environment:

    Run on: affected node
    grep '^ldap_search_base' /etc/sssd/sssd.conf
    getent passwd <ldap-user>
    sssctl user-show <ldap-user>
    
    Run on: OIM
    podman exec omnia_auth ldapsearch -x -H ldap://127.0.0.1 \
      -b '<configured-search-base>' '(uid=<ldap-user>)' \
      objectClass uidNumber gidNumber homeDirectory loginShell
    
  2. Ensure the entry has the posixAccount attributes required by the cluster. Use the site's approved LDAP administration workflow to repair the entry; do not copy a fixed administrator DN from an example.

  3. If the server entry is correct, clear only SSSD's cached entry and restart SSSD:

    Run on: affected node
    sss_cache -u <ldap-user>
    systemctl restart sssd
    getent passwd <ldap-user>
    

User authentication fails on a cluster node

Symptom

The LDAP identity resolves, but SSH, su, or PAM authentication fails.

Resolution
  1. Verify SSSD and the active authselect profile:

    Run on: affected node
    systemctl status sssd --no-pager
    authselect current
    journalctl -u sssd -b -n 200 --no-pager
    
  2. Confirm the node can reach the URI and port in /etc/sssd/sssd.conf. TLS in security_config.yml uses StartTLS on port 389; SSL uses LDAPS on port 636.

  3. Confirm the generated profile enables SSSD and home-directory creation:

    Run on: affected node
    authselect select sssd with-mkhomedir --force
    systemctl enable --now sssd
    

authconfig and nslcd are legacy mechanisms and are not used by the current Orchestrator templates. 4. When generated settings are wrong, correct the Orchestrator input and rerun precheck and provision. Avoid maintaining a manual SSSD configuration that the next provisioning run will replace.

LDAP TLS certificate error

Symptom

SSSD or ldapsearch reports an untrusted, expired, or mismatched certificate.

Resolution
  1. Inspect the certificate generated for omnia_auth on the OIM:

    Run on: OIM
    source /etc/profile.d/omnia-env.sh
    openssl x509 -in "$OMNIA_DATA_PATH/auth/tls_certs/ldapserver.crt" \
      -noout -subject -issuer -dates -ext subjectAltName
    
  2. Inspect the provisioned client copy and SSSD reference:

    Run on: affected node
    openssl x509 -in /etc/openldap/certs/ldapserver.crt \
      -noout -subject -issuer -dates -ext subjectAltName
    grep -E 'ldap_uri|ldap_tls_cacert|ldap_tls_reqcert' /etc/sssd/sssd.conf
    
  3. If the client copy differs, rerun Orchestrator provisioning so the generated metadata copies the current certificate and configuration. Do not use the obsolete /opt/omnia/omnia/openldap/certs path.

  4. Restart SSSD and verify identity lookup:

    Run on: affected node
    systemctl restart sssd
    getent passwd <ldap-user>
    

SSH reports a changed host key

This is an SSH trust failure, not an LDAP authentication failure. Confirm that the host was intentionally reprovisioned before removing its old key:

Run on: SSH client
ssh-keygen -R <node-hostname-or-IP>
ssh-keyscan <node-hostname-or-IP> | ssh-keygen -lf -

Verify the displayed fingerprint through the site's trusted inventory before adding it to known_hosts.

See Deploy OpenLDAP for the supported configuration and deployment workflow.