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:
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:
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
-
Inspect the service and container logs:
Run on: OIMsystemctl status omnia_auth --no-pager journalctl -u omnia_auth -b -n 200 --no-pager podman logs --tail 200 omnia_auth -
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.
-
Rerun the supported preparation and validation flows:
Run on: OIM hostcd <OMNIA_SOURCE_PATH>/src/main ./omnia.sh --run orchestrator --tags prepare ./omnia.sh --run orchestrator --tags validate-deploymentRun on: OIM hostsource /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
-
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 nodegrep '^ldap_search_base' /etc/sssd/sssd.conf getent passwd <ldap-user> sssctl user-show <ldap-user>Run on: OIMpodman exec omnia_auth ldapsearch -x -H ldap://127.0.0.1 \ -b '<configured-search-base>' '(uid=<ldap-user>)' \ objectClass uidNumber gidNumber homeDirectory loginShell -
Ensure the entry has the
posixAccountattributes 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. -
If the server entry is correct, clear only SSSD's cached entry and restart SSSD:
Run on: affected nodesss_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
-
Verify SSSD and the active authselect profile:
Run on: affected nodesystemctl status sssd --no-pager authselect current journalctl -u sssd -b -n 200 --no-pager -
Confirm the node can reach the URI and port in
/etc/sssd/sssd.conf.TLSinsecurity_config.ymluses StartTLS on port 389;SSLuses LDAPS on port 636. -
Confirm the generated profile enables SSSD and home-directory creation:
Run on: affected nodeauthselect 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
-
Inspect the certificate generated for
omnia_authon the OIM:Run on: OIMsource /etc/profile.d/omnia-env.sh openssl x509 -in "$OMNIA_DATA_PATH/auth/tls_certs/ldapserver.crt" \ -noout -subject -issuer -dates -ext subjectAltName -
Inspect the provisioned client copy and SSSD reference:
Run on: affected nodeopenssl 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 -
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/certspath. -
Restart SSSD and verify identity lookup:
Run on: affected nodesystemctl 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:
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.