Two likely possibilities:
1. Maybe because /home is not mounted.
2. Maybe because users have been configured by Bright Cluster Manager to prevent ssh access.
So check these first.
/home not mounted
By default, a regular user requires
/home to be mounted for the passwordless ssh logins, because that’s where the authorized keys are kept.
A root user keeps their authorized keys in
/root, which is not normally mounted separately.
The Admin Manual covers this in the section on “Logins Using ssh”
Prevention of ssh access with BCM
Preventing ssh access is typically used in systems with workload managers, to prevent regular users using up resources outside of workload managers. The Admin Manual covers this in section on “Forcing Jobs To Run In A Workload Management System”
A less common possibility — is LDAP broken?
If LDAP for the regular nodes is broken in some way, it can cause the user ids to be unknown and also lead to this behavior. A broken LDAP setup can happen, for example, when creating a software image from scratch with create-image, and forgetting an LDAP package.