Setup
- Endpoint Protector virtual appliance on VMware ESXi
- Keycloak as SAML 2.0 Identity Provider, configured via the Okta profile
- Internal test environment
What happens
SSO login fails, and the failure mode changes depending on the appliance clock:
- Appliance clock off by 1h42m: EPP shows the regular SSO error page (“There has been an issue with your Single Sign On Process”).
- Appliance clock synchronised to the IdP (0 seconds difference): EPP returns “Oops! An Error Occurred - The server returned a 500 Internal Server Error”.
Both reproduced twice, each matching the state of the clock.
What I already checked
- The clock skew is fixed. The default NTP server (pool.ntp.org) is not reachable from this network segment, so I replaced it with an internal NTP server under Appliance > Server Maintenance > Time Zone. Both clocks now match to the second, verified in UTC on both sides.
- On the Keycloak side, a client session for the Endpoint Protector service is created on every attempt. So the IdP authenticates the user and issues the assertion - the failure happens inside EPP after the assertion is received.
- Identity and attribute matching: four different configurations were tested earlier, all failed identically while the clock was still wrong.
Questions
- Has anyone seen a 500 on the SSO callback with a correctly synchronised clock?
- Where does the appliance log SAML processing errors when SSH access is not available?