[389-users] Change name of server, admin-server no longer works

夜神 岩男 supergiantpotato at yahoo.co.jp
Fri Jul 29 20:41:59 UTC 2011

On 07/30/2011 05:17 AM, Techie wrote:
> 2011/7/29 夜神 岩男<supergiantpotato at yahoo.co.jp>:
>> On 07/29/2011 04:34 PM, Techie wrote:
>>> Hello,
>>> We were required to change the hostname of our LDAP server running
>>> 389-DS. Since that time the LDAP server runs fine but the admin server
>>> does not authenticate login any longer, meaning i cannot log into the
>>> admin server. What do I need to do to fix the admin server and change
>>> all references from the old host name to the new host name.
>> Just for clarity, what does "admin server" mean:
> The admin-server is the Java front end/interface that allows you to
> admin the server via http.
> So you connect like..
> http://myserver:9080
> Then you can admin the LDAP instance via GUI.

> LDAP works fine.. It is the Java admin-server that is broken. It is
> broken because hte references under the config files under
> /etc/dirsrv/admin-serv are pointing to the incorrect host name. I am
> not sure if me simply changing all references to the new hostname will
> fix it.

Fixing the hostname references is part of it, and if you are using 
certificates specific to the admin-server to authenticate then they need 
to be updated/replaced as well to avoid things like instance/realm or 
nss hostname check problems.

The config files should contain lots of references to the old hostname 
(unless a magical script fixed them when you weren't looking), and those 
must be changed. Don't forget to look places like nss.conf, and weirder 
areas like filnames of auth keys (and make sure to check silly spots 
like hosts.conf to make sure NetworkManager or whatever didn't append 
the new hostname in there somewhere (like an unused IPv6 line), or mix 
and match old and new hostnames, as this can break random authentication 
things related to Kerberos and NSS). Some files have hostname info 
tagged at the end of them, and things that point to them must be lined up.

I would start by walking myself back through manual setup steps as if I 
were setting up admin-server on a new system to make sure I didn't miss 
anything and then recreating my authentication keys if necessary.

Fixing a partially broken authentication setup *sucks*. In situations 
like that if the machine isn't the sole server (a slave is out there 
somewhere), I'll just re-install the server packages to make sure 
nothing is missed and then replicate back from the slave or a backup 
because re-setting nitpicky manual setups without doing them 100% from 
the beginning can be a real pain.


More information about the 389-users mailing list