Monday, August 26, 2024

[389-users] Re: [EXT] Re: Password policies and replication service accounts


On 8/26/24 6:33 PM, Darby, Tim - (tdarby) wrote:
I know about the replication agreements, cn=replication manager, and all that, but...
Sorry I wasn't trying to be rude or condescending - I just wanted to make sure it was setup right because "Inappropriate auth" is usually when a password is not provided (or doesn't exist in the entry?).
In the process of moving my old instances to 2.5, I decided to use lib389 as much as possible to script my replicas, replication agreements, replication accounts, etc. lib389, as far as I can tell, only creates replication service accounts in ou=Services. As such, I assumed that the old system, where everything was in cn=config, was no longer the recommendation.
That is still fine, and the CLI handles both.  The local(subtree/user) polices are just for fine-grained control, and these policy configurations replicate which is nice (set it on one server and you're good to go).  Those are its pros, but the global password policy under cn=config is perfectly fine as well.
It seems to me that having the replication accounts in an OU makes them susceptible to password polices, which I do believe was causing the issues I was seeing.
If the service account does not yet exist on the consumer, then yes it will cause issues (error 32/49 on bind).  We actually have bootstrap settings to help in these scenario (but I'm not sure that's what your issue is).
If you're saying that having it all in cn=config is still fine, I will happily revert to that.

You shouldn't have to though.  I think there is just some conflict going on (or misconfiguration).  In the 2.x versions there is a new password policy error logging level (1048576),  It will tell you why a bind/password update failure happens and by what policy. 



On subtree policies, if that's the expected behavior of dsconf, which I find rather unhelpful to be honest, then it's all good. It just feels like there out to be something in the system that you can point at any given ou and ask "what's the policy that's active on this ou?"

Ahh yeah, that would be a nice improvement to the CLI/UI, and definitely doable!


Back to replication.  If you say password policy is the reason for the error 48 -> inappropriate auth, then I disagree that is the root cause.  I think turning on replication logging (level 8192) will help point us where to look.  Password policy issues usually results in an error 49 or 53.  That's why I think it's not password policy related.


Either way we need more info.  It might be best to turn on both of those error logging levels (combine them 1048576+8192=1056768).  Just set it back to 0 when you're done reproducing the issue (do this on both supplier and consumer).  Then send us the log clips please.


Thanks,

Mark


Tim Darby

From: Mark Reynolds <mareynol@redhat.com>
Sent: Monday, August 26, 2024 12:13
To: General discussion list for the 389 Directory server project. <389-users@lists.fedoraproject.org>; Darby, Tim - (tdarby) <tdarby@arizona.edu>
Subject: Re: [389-users] Re: [EXT] Re: Password policies and replication service accounts
 

External Email




On 8/26/24 1:18 PM, Darby, Tim - (tdarby) wrote:
Thanks, I'm reviewing what I did to see where this went wrong. I did supply credentials for the service accounts.

I was referring to the entry you use in the replication agreement (typically cn=replication manager, cn=config):


Supplier:


dn: cn=replication manager,cn=config
objectClass: top
objectClass: inetUser
objectClass: netscapeServer
objectClass: nsAccount
cn: replication manager
uid: replication manager
userPassword:: e1BCS0RGMi1TSEE1MTJ9MTAwMDAkQ0c5N08xK2crMWZyU0czQkdhcjN6WTNqM2Y
 2eHEvMHEkdXlPTnhmdW43RUEycVBFcmtKMHdXVGtSUXNpL3VBbDVpQTNySHJkRFJZejZGZTdjcHpi
 SHYzRnRQNlJ0Nnl4a1dvTFJ6clJ3a1lnYk9mMmo1OSsyZHc9PQ==

dn: cn=darby,cn=replica,cn=dc\3Dexample\2Cdc\3Dcom,cn=mapping tree,cn=config
objectClass: top
objectClass: nsds5replicationagreement
cn: darby
nsDS5ReplicaRoot: dc=example,dc=com
description: darby
nsDS5ReplicaHost: localhost
nsDS5ReplicaPort: 389
nsDS5ReplicaBindMethod: simple
nsDS5ReplicaTransportInfo: LDAP
nsDS5ReplicaBindDN: cn=replication manager,cn=config
nsDS5ReplicaCredentials: {AES-TUhNR0NTcUdTSWIzRFFFRkRUQm1NRVVHQ1NxR1NJYjNEUUVG
 RERBNEJDUmxObUk0WXpjM1l5MHdaVE5rTXpZNA0KTnkxaE9XSmhORGRoT0MwMk1ESmpNV014TUFBQ
 0FRSUNBU0F3Q2dZSUtvWklodmNOQWdjd0hRWUpZSVpJQVdVRA0KQkFFcUJCQUVqZlE0RzhhaHF5Rz
 dUN0F3Mmk3WQ==}14bMwBr/FBN6L1kPTBuyRA==


On the consumer side you must specify the bind that that can perform replication updates in the replica configuration entry:


dn: cn=replica,cn=dc\3Dexample\2Cdc\3Dcom,cn=mapping tree,cn=config
objectClass: top
objectClass: nsds5Replica
cn: replica
nsDS5ReplicaRoot: dc=example,dc=com
nsDS5ReplicaBindDN: cn=replication manager,cn=config

...

...



Question about inheritance of subtree password policies: How do you see that it has been applied to a subtree of a subtree? dsconf claims that there is no subtree policy on ou=test,ou=accounts:

> dsconf -D "cn=Directory Manager" -w "xxxx" ldap://localhost:3389 localpwp list
ou=accounts,ou=etc etc (subtree policy)

> dsconf -D "cn=Directory Manager" -w "xxxx" ldap://localhost:3389 localpwp get "ou=test,ou=accounts,ou=etc etc"
Error: No password policy was found for this entry

Right, so anything under "ou=accounts,ou=etc" should have that local policy applied to it.  It will only be listed under the the original subtree.  So this all looks correct.  Please provide your password policy settings and describe how it's not working as you expect:


    # dsconf slapd-INSTANCE localpwp get "ou=accounts,ou=etc"



Thanks,
Mark


Tim Darby
Systems Integration and Architecture | The University of Arizona | tdarby@arizona.edu | they/he

From: Mark Reynolds <mareynol@redhat.com>
Sent: Monday, August 26, 2024 07:14
To: General discussion list for the 389 Directory server project. <389-users@lists.fedoraproject.org>; Darby, Tim - (tdarby) <tdarby@arizona.edu>
Subject: [EXT] Re: [389-users] Password policies and replication service accounts
 
External Email


On 8/19/24 11:05 AM, tdarby@arizona.edu wrote:
> I encountered a problem with replication service accounts in the process of moving my one remaining very old (1.9.x) 389ds system to the new container. This is likely a misunderstanding on my part about how password policies work, so I'd appreciate any insights on this.
>
> This system has two MMR instances and an ou=Accounts containing thousands of user accounts. It uses a global password policy for the user accounts (there's no subtree policy on ou=Accounts). As I did with a previous migration to 3.x, I created a new ou, ou=Services,ou=Accounts, to hold the service accounts for replication. When I tried to do the initial replication, it failed because the source instance couldn't authenticate to the destination instance. I was seeing weird messages like "inappropriate authentication", etc.
>
> It occurred to me that maybe the issue had to do with the fact that this system had a global policy set whereas the previous system was not using a global policy. I tried removing the global policy and adding a subtree policy instead to ou=Accounts and that solved the problem. So, questions:
>
> - Is there a way to have a global password policy but not have it apply to a particular ou?

The global password policy under cn=config applies to all entries in the
database.  Then subtree policies (or fine-grained policies) can
over-rule the global policy.  If you want the global and subtree
policies to blend together then you must set the polices to inherit the
global policy:

https://docs.redhat.com/en/documentation/red_hat_directory_server/11/html/configuration_command_and_file_reference/core_server_configuration_reference#cnconfig-nsslapd_pwpolicy_inherit_global_Inherit_Global_Password_Policy

But, if you want the subtree policy to completely bypass the global
policy then do NOT set that attribute from the doc.

> - It appears that setting a subtree policy on an ou (ou=Accounts), does not inherit to a subtree of that tree (ou=Services). Is that right?
It should apply to all entries under the subtree policy, if not it's a
bug.  But we have tests for this so it should definitely be working in
newer versions of 389.
> - It's not clear to me what actually causes the global policy to be active. Does it become active simply by changing any of its password attributes in cn=config?

Yes, but like I said subtree policies overrule it.  All changes to
global/fine-grain polices take effect immediately.


Now going back to your replication issue, the password policy should not
impact replication.  Inappropriate auth means a password/credential was
not provided.  It's possible the consumer does not have a replication
manager defined, or you left out the credentials attribute in the
agreement.  Either way that's a replication config issue (agreement or
replica config) and unrelated to password policy.

HTH,

Mark

--
Identity Management Development Team

--   Identity Management Development Team

--   Identity Management Development Team

No comments:

Post a Comment