[HDFS-13520] fuse_dfs to support keytab based login

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 2.6.0

Environment

Hadoop 2.6/3.0, Kerberized, fuse_dfs

• Assignee: Unassigned
• Reporter: Wei-Chiu Chuang
• Created: 02/May/18 18:14
• Updated: 02/May/18 18:17

Description

It looks like the current fuse_dfs implementation supports login using current kerberos credential. If the tgt expires, it fails with the following error:

hdfsBuilderConnect(forceNewInstance=1, nn=hdfs://ns1, port=0, kerbTicketCachePath=/tmp/krb5cc_2000, userName=systest) error:
LoginException: Unable to obtain Principal Name for authentication org.apache.hadoop.security.KerberosAuthException: failure to login: for user: systest using ticket cache file: /tmp/krb5cc_2000 javax.security.auth.login.LoginException: Unable to obtain Principal Name for authentication
at org.apache.hadoop.security.UserGroupInformation.getUGIFromTicketCache(UserGroupInformation.java:807)
at org.apache.hadoop.security.UserGroupInformation.getBestUGI(UserGroupInformation.java:742)
at org.apache.hadoop.fs.FileSystem.newInstance(FileSystem.java:404)
Caused by: javax.security.auth.login.LoginException: Unable to obtain Principal Name for authentication
at com.sun.security.auth.module.Krb5LoginModule.promptForName(Krb5LoginModule.java:841)
at com.sun.security.auth.module.Krb5LoginModule.attemptAuthentication(Krb5LoginModule.java:704)
at com.sun.security.auth.module.Krb5LoginModule.login(Krb5LoginModule.java:617)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at javax.security.auth.login.LoginContext.invoke(LoginContext.java:755)
at javax.security.auth.login.LoginContext.access$000(LoginContext.java:195)
at javax.security.auth.login.LoginContext$4.run(LoginContext.java:682)
at javax.security.auth.login.LoginContext$4.run(LoginContext.java:680)
at java.security.AccessController.doPrivileged(Native Method)
at javax.security.auth.login.LoginContext.invokePriv(LoginContext.java:680)
at javax.security.auth.login.LoginContext.login(LoginContext.java:587)
at org.apache.hadoop.security.UserGroupInformation.getUGIFromTicketCache(UserGroupInformation.java:788)
... 2 more

This is reproducible easily in a test cluster with an extremely short ticket life time (e.g. 1 minute)

Note: HDFS-3608 addresses a similar issue, but in this case, since the ticket cache file itself does not change, fuse couldn't detect & update.

It looks like it should call UserGroupInformation#loginFromKeytab() in the beginning, similar to how balancer supports keytab based login (HDFS-9804). Thanks xiaochen for the idea.

A quick workaround would have a crontab job that periodically renew the kerberos ticket with a keytab, say every 8 hours.

Issue Links

• relates to: HDFS-3608 fuse_dfs: detect changes in UID ticket cache
• relates to: HDFS-9804 Allow long-running Balancer to login with keytab