Hi all, I''ve been all day trying to get simple single master to one consumer going on a pair of 1.0.4 FDS systems and I can''t get past the authentication credentials. I''ve gone over this 20 times today from scratch, and it won''t go. I''ve even redone my procedures on my test boxes and they work fine. Both the replication wizard and the consumer initialization fail (if I force the wizard to accept and go on). There is no firewall issue and tcpdump and ldapsearch gets to the consumer machine. Consumer is RHEL 4. Here''s my LDIF''s I used. Can I use ldapsearch to test binding to this netry to try an debug what''s up? On an aside, the Redhat/Fedora documents for adding this entry are very vague and don''t have any information about most of these attributes. It didn''t appear one could even get this working *without* using LDIF files. Anyway, any help would be great. Thanks. dn: cn=replica, cn="dc=acme,dc=com", cn=mapping tree, cn=config changetype: add objectClass: nsDS5Replica objectClass: top cn: replica nsDS5ReplicaBindDN: cn=Replication Manager, cn=config nsDS5ReplicaRoot: dc=acme,dc=com nsDS5Flags: 0 nsDS5ReplicaType: 2 nsDS5ReplicaId: 1 dn: cn=Replication Manager, cn=config changetype: add cn: Replication Manager sn: replication objectClass: top objectClass: person userPassword: xxxxx -- - Kyle --------------------------------------------- kylet@panix.com http://www.panix.com/~kylet ---------------------------------------------
koniczynek
2006-Nov-25 07:21 UTC
Re: [Fedora-directory-users] Replication credentials issue
Kyle Tucker, dnia 2006-11-25 02:39 napisal:> dn: cn=Replication Manager, cn=config > changetype: add > cn: Replication Manager > sn: replication > objectClass: top > objectClass: person > userPassword: xxxxx''userPassword'' field should be clear text. So considering that you entered ''xxxxx'' in this example, in the wizard you should provide ''xxxxx'' as a password. Easiest way to do it: stop your replica, edit dse.ldif (can be found in ./slapd-hostname/config/) and type password in there in the userPassword. Start replica and run replication agreement wizard on the master. Initialization should go on without any problems. -- email/xmpp: koniczynek@uaznia.net
Kyle Tucker
2006-Nov-25 13:28 UTC
Re: [Fedora-directory-users] Replication credentials issue
> Kyle Tucker, dnia 2006-11-25 02:39 napisal: > > dn: cn=Replication Manager, cn=config > > changetype: add > > cn: Replication Manager > > sn: replication > > objectClass: top > > objectClass: person > > userPassword: xxxxx > ''userPassword'' field should be clear text. So considering that you > entered ''xxxxx'' in this example, in the wizard you should provide > ''xxxxx'' as a password. Easiest way to do it: stop your replica, edit > dse.ldif (can be found in ./slapd-hostname/config/) and type password in > there in the userPassword. Start replica and run replication agreement > wizard on the master. Initialization should go on without any problems.I put xxxxx in this just as an example. Oddly, on my working test servers (1.0.3 on FC5), I used a {SSHA} password in my LDIF and it worked without a hitch all 3 times I''ve set replication there. I did put in clear text in the non-working production sets. If it''s of any help, here''s the entries in the access logs, although I''d guess there''s nothing revelational here. [24/Nov/2006:17:26:56 -0800] conn=23 fd=64 slot=64 connection from 10.1.100.186 to 10.1.109.203 [24/Nov/2006:17:26:56 -0800] conn=23 op=0 BIND dn="cn=Replication Manager,cn=config" method=128 version=3 [24/Nov/2006:17:26:56 -0800] conn=23 op=0 RESULT err=49 tag=97 nentries=0 etime=0 [24/Nov/2006:17:26:56 -0800] conn=23 op=1 UNBIND [24/Nov/2006:17:26:56 -0800] conn=23 op=1 fd=64 closed - U1 -- - Kyle --------------------------------------------- kylet@panix.com http://www.panix.com/~kylet ---------------------------------------------
Kyle Tucker
2006-Nov-25 13:41 UTC
Re: [Fedora-directory-users] Replication credentials issue
> > > userPassword: xxxxx > > ''userPassword'' field should be clear text. So considering that you > > entered ''xxxxx'' in this example, in the wizard you should provide > > ''xxxxx'' as a password. Easiest way to do it: stop your replica, edit > > dse.ldif (can be found in ./slapd-hostname/config/) and type password in > > there in the userPassword. Start replica and run replication agreement > > wizard on the master. Initialization should go on without any problems.I stopped the service, edited the password in clear in userPassword field, reinput the password on the master and same errors. The error from the initialize consumer action is: The consumer initialization has unsuccessfully completed. The error received by the replica is: ''49 - LDAP error: Invalid credentials''. Corresponding log entry is the same: [25/Nov/2006:05:37:17 -0800] conn=1 op=0 BIND dn="cn=Replication Manager,cn=config" method=128 version=3 [25/Nov/2006:05:37:17 -0800] conn=1 op=0 RESULT err=49 tag=97 nentries=0 etime=0 [25/Nov/2006:05:37:17 -0800] conn=1 op=1 UNBIND [25/Nov/2006:05:37:17 -0800] conn=1 op=1 fd=64 closed - U1 So how to debug this is the next step it seems. Thanks. -- - Kyle --------------------------------------------- kylet@panix.com http://www.panix.com/~kylet ---------------------------------------------
Kyle Tucker
2006-Nov-25 13:58 UTC
Re: [Fedora-directory-users] Replication credentials issue
> I stopped the service, edited the password in clear in userPassword > field, reinput the password on the master and same errors. The error > from the initialize consumer action is:For grins, I stopped the master as well, edited its dse.ldif and changed it to clear (it was in DES method) and voila - it all took off and synched up. I checked my working test master and consumer and they were in DES and SSHA respectively, again always working from the onset. I''ll leave it to the developers to take anything from this. Thanks for the pointer to dse.ldif. -- - Kyle --------------------------------------------- kylet@panix.com http://www.panix.com/~kylet ---------------------------------------------
Richard Megginson
2006-Nov-27 16:55 UTC
Re: [Fedora-directory-users] Replication credentials issue
Kyle Tucker wrote:>> I stopped the service, edited the password in clear in userPassword >> field, reinput the password on the master and same errors. The error >> from the initialize consumer action is: >> > > For grins, I stopped the master as well, edited its dse.ldif and > changed it to clear (it was in DES method) and voila - it all took > off and synched up. I checked my working test master and consumer > and they were in DES and SSHA respectively, again always working > from the onset. I''ll leave it to the developers to take anything from > this. Thanks for the pointer to dse.ldif. >The consumer should have the cn=Repl Manager user with userPassword as an SSHA hash (or some other secure hash), not cleartext. The supplier should store the repl manager credentials with the {DES} reversible password encryption type so that it can send the clear text password to the consumer in the BIND request (as is done in the normal LDAP simple BIND request). You can always test this by using the ldapsearch command line tool to attempt to bind using -D "cn=replication manager,cn=config" and the password to the consumer to test the bind and credentials.
Kyle Tucker
2006-Nov-27 17:57 UTC
Re: [Fedora-directory-users] Replication credentials issue
> Kyle Tucker wrote: > >> I stopped the service, edited the password in clear in userPassword > >> field, reinput the password on the master and same errors. The error > >> from the initialize consumer action is: > >> > > > > For grins, I stopped the master as well, edited its dse.ldif and > > changed it to clear (it was in DES method) and voila - it all took > > off and synched up. I checked my working test master and consumer > > and they were in DES and SSHA respectively, again always working > > from the onset. I''ll leave it to the developers to take anything from > > this. Thanks for the pointer to dse.ldif. > > > The consumer should have the cn=Repl Manager user with userPassword as > an SSHA hash (or some other secure hash), not cleartext. The supplier > should store the repl manager credentials with the {DES} reversible > password encryption type so that it can send the clear text password to > the consumer in the BIND request (as is done in the normal LDAP simple > BIND request). You can always test this by using the ldapsearch command > line tool to attempt to bind using -D "cn=replication manager,cn=config" > and the password to the consumer to test the bind and credentials.Yes, but it wouldn''t work in this configuration using DES->SSHA with 1.0.4 on RHEL, whereas it did in several tests on 1.0.3 on FC5. It wouldn''t even work DES->clear. I did not try clear->SSHA. I have to set up 2 more consumers, so I will try all possible combinations when I do those and follow up. -- - Kyle --------------------------------------------- kylet@panix.com http://www.panix.com/~kylet ---------------------------------------------