[ 
https://issues.apache.org/jira/browse/AMBARI-6432?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15185813#comment-15185813
 ] 

Bolke de Bruin commented on AMBARI-6432:
----------------------------------------

[~u39kun] Thanks that will help.

In the meantime I also hit a bug on which I need some guidance how to fix it. 
FreeIPA does not support uppercase user principal names. If the cluster name is 
in uppercase a test identity will be generated with "myname@REALM" . For the 
tests to pass I need it to be "myname@REALM".

What would be the best way to fix this?

1) Generate the test identity in lowercase (where is this generated? I could 
not find it yet)
2) Ask the user at step 1 to verify the cluster name is in lower case
3) adjust service_check.py to convert to lowercase if a user principal is 
encountered
4) use auth_to_local rules and apply these to the test identity

In my opinion option 1 seems the best option. It would affect other Kerberos 
providers as well, but as it is only the test identity I would say it would not 
matter.

Please advice 

> FreeIPA Support in Ambari
> -------------------------
>
>                 Key: AMBARI-6432
>                 URL: https://issues.apache.org/jira/browse/AMBARI-6432
>             Project: Ambari
>          Issue Type: Improvement
>          Components: ambari-server
>    Affects Versions: trunk
>            Reporter: jay vyas
>            Assignee: Bolke de Bruin
>             Fix For: 2.4.0
>
>         Attachments: AMBARI-6432-FreeIPA.patch, AMBARI-6432.patch, 
> AMBARI-6432.trunk.v1.patch, AMBARI-6432.trunk.v2.patch, 
> AMBARI-6432.trunk.v3.patch, AMBARI-6432.trunk.v4.patch, 
> AMBARI-6432.trunk.v5.patch, AMBARI-6432.trunk.v5.patch, ipa-patch-v0.5.patch
>
>
> FreeIPA Is a powerful tool for unifying identity, kerberos credentials, 
> across a cluster.
> A great value add for ambari would be to provide support for using FreeIPA to 
> kerberize services.  This would allow for 
> 1) better HCFS interoperability, because first class GID/UID is critical for 
> certain file systems (GlusterFS, Lustre, and any other file system which uses 
> kernel / FUSE apis for determining identity)
> 2) better enterprise interoperability.  Because of the fact that FreeIPA 
> makes it easy to interop with different identity solutions (like active 
> directory), it would make ambari easier to adopt for various enterprises.
> 3) broadens ambaris scope.  Now ambari could also allow people to setup the 
> users of their clusters, and at least some of the security features of their 
> clusters, all from one interface (no more manual handling of TGTs and such - 
> it could all be done quite easily via the ambari UI which could make calls to 
> underlying FreeIPA clients).



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to