Common Kerberos exceptionsedit

Symptoms:

  • User authentication fails due to either GSS negotiation failure or a service login failure (either on the server or in the Elasticsearch http client). Some of the common exceptions are listed below with some tips to help resolve them.

Resolution:

Failure unspecified at GSS-API level (Mechanism level: Checksum failed)

When you see this error message on the HTTP client side, then it may be related to an incorrect password.

When you see this error message in the Elasticsearch server logs, then it may be related to the Elasticsearch service keytab. The keytab file is present but it failed to log in as the user. Please check the keytab expiry. Also check whether the keytab contain up-to-date credentials; if not, replace them.

You can use tools like klist or ktab to list principals inside the keytab and validate them. You can use kinit to see if you can acquire initial tickets using the keytab. Please check the tools and their documentation in your Kerberos environment.

Kerberos depends on proper hostname resolution, so please check your DNS infrastructure. Incorrect DNS setup, DNS SRV records or configuration for KDC servers in krb5.conf can cause problems with hostname resolution.

Failure unspecified at GSS-API level (Mechanism level: Request is a replay (34)) , Failure unspecified at GSS-API level (Mechanism level: Clock skew too great (37))
To prevent replay attacks, Kerberos V5 sets a maximum tolerance for computer clock synchronization and it is typically 5 minutes. Please check whether the time on the machines within the domain is in sync.

As Kerberos logs are often cryptic in nature and many things can go wrong as it depends on external services like DNS and NTP. You might have to enable additional debug logs to determine the root cause of the issue.

Elasticsearch uses a JAAS (Java Authentication and Authorization Service) Kerberos login module to provide Kerberos support. To enable debug logs on Elasticsearch for the login module use following Kerberos realm setting:

xpack.security.authc.realms.<realm-name>.krb.debug: true

For detailed information, see Kerberos realm settings.

Sometimes you may need to go deeper to understand the problem during SPNEGO GSS context negotiation or look at the Kerberos message exchange. To enable Kerberos/SPNEGO debug logging on JVM, add following JVM system properties:

-Dsun.security.krb5.debug=true

-Dsun.security.spnego.debug=true

For more information about JVM system properties, see configuring JVM options.