-
Notifications
You must be signed in to change notification settings - Fork 138
LDAP Profile Storage
Dogtag supports adding and modifying CA profiles. New profiles are created on the filesystem alongside the default profiles installed by Dogtag, and updates to profiles are performed in-place. There is no mechanism to replicate profiles.
It is now required to implement support for CA profiles in FreeIPA. Because FreeIPA is typically deployed in a multi-master replication configuration, with replication agreements between the LDAP databases on separate hosts, and because changes to profiles will also need to be replicated, this document proposes a design for LDAP-based profile storage in Dogtag.
It is furthermore noted that according to alee, LDAP profile storage and replication has been "on the wishlist" for a while.
This feature is slated for version 10.3, or possibly a future 10.2.x release.
Some related tickets that it may make sense to attack whilst implementing this proposal:
There is an RFE for Ticket 1631 - Shared CA Enrollment Profiles, i.e., shared across instances. The design should be mindful of this RFE.
A FreeIPA user adds a new profile for a certain use case, e.g. a client certificate with a certain application-specific X.509 extension (the interface for defining/importing profiles in FreeIPA is beyond scope of this proposal). As a result of this action, the profile should be available on all replicas.
Similarly, modifying a profile should result in the modification being effected on all replicas.
Linux (Fedora, RHEL, Debian).
The essence of the design, as explained by alee is:
-
Continue to distribute the system profiles in flat files. These files will be read and stored in LDAP when an instance is created.
-
All profiles for an instance should live in LDAP. This makes it simple - no need to check to see if a profile is in LDAP or files or both, and which has priority etc. Tools will be provided to manage/create/delete profiles etc.
-
Updates to system profile files will not affect the existing LDAP profiles. We can provide update scripts or manual instructions for admins to run when they opt to do so. This will be for behavioral changes. If IPA has changes to their profiles, they can apply through the ldap update mechanisms they have in place.
-
Structural changes will be done using upgrade scripts using the database upgrade mechanism. This framework is something that we had planned to do already in 10.3. We already have a model on how to do this in our current upgrade framework.
Because profiles are currently stored as configuration for a particular CA subsystem, it follows that LDAP profiles will be stored as attributes of a particular CA or sub-CA subsystem. This will be simpler to implement and ensure that deployments prior to, or not using the Top-level Tree capability can take advantage of LDAP profile storage.
The existing profile registry stores the path to the profile configuration file and a reference to the enrollment implementation. For LDAP profiles, the data that would be stored in the profile configuration file will be stored according to the following schema, which will be defined in schema.ldif
.
The certProfile
terminology is used where possible to disambiguate certificate profiles from TPS token profiles.
The classId
attribute is a Directory String that stores the the enrollment class identifier:
dn: cn=schema changetype: modify add: attributeTypes attributeTypes: ( classId-oid NAME 'classId' DESC 'CMS defined attribute' SYNTAX 1.3.6.1.4.1.1466.115.121.1.15 X-ORIGIN 'user defined' )
The certProfileConfig
attribute is an Octet String that stores the profile configuration (the same format as is currently stored in files):
dn: cn=schema changetype: modify add: attributeTypes attributeTypes: ( certProfileConfig-oid NAME 'certProfileConfig' DESC 'CMS defined attribute' SYNTAX 1.3.6.1.4.1.1466.115.121.1.40 X-ORIGIN 'user defined' )
The certProfile
object class defines the complete profile record:
dn: cn=schema changetype: modify add: objectClasses objectClasses: ( certProfile-oid NAME 'certProfile' DESC 'CMS defined class' SUP top STRUCTURAL MUST cn MAY ( classId $ certProfileConfig ) X-ORIGIN 'user defined' )
Profiles will be stored under a new OU:
dn: ou=certificateProfiles,ou=ca,{rootSuffix} objectClass: top objectClass: organizationalUnit ou: certificateProfiles
According to the above schema, LDAP-based profile records will look like:
dn: cn=<certProfileId>,ou=certificateProfiles,ou=ca,{rootSuffix} objectClass: top objectClass: certProfile cn: <certProfileId> classId: <classId> certProfileConfig: <octet string>
The ProfileSubsystem
will be changed to use the LDAP database as its data store instead of the filesystem. This should require no significant changes to its public API.
Currently, profiles are read at startup. This means that we need some mechanism to trigger the refreshing of the profiles (without restart) when changes made on other clones are replicated to the local database.
A persistent LDAP search will be executed in its own thread to monitor for changes that occur beneath the certificateProfiles
OU and update the ProfileSubsystem
accordingly. The entire implementation is contained within the ProfileSubsystem
class. Specifically, the presistent query is executed by the run()
method of the Monitor
inner class, which extends Thread
.
The REST API should not require any significant changes. Any changes that are required should be reflected in the Python API.
Currently, only Administrators can create, modify or delete profiles. No changes to this access control are proposed.
(alee) Dogtag uses its own system of ACLs, which are enforced on the servlet level. Creating/changing profiles are done through servlets and access controls are enforced there. This allows us to do complex things like requiring agents to disable a profile before an admin can edit it.
Users do not access the dogtag internal database directly. Rather, the database is only accessed via a special system user that performs operations on behalf of the server.
In any case, this mechanism is not going to change. We will keep the same Dogtag servlet ACLs, so the behavior will be the same.
Editing of file-based profiles has until now been a simple matter of editing the file and restarting Dogtag so that profile changes take effect. With profiles now to be stored in LDAP, new mechanisms are needed to edit profiles.
The --raw
flag will be added to the existing CLI commands (ca-profile-show
, ca-profile-add
and ca-profile-mod
) for working with profiles in the "raw" format, rather than the XML format.
The pki ca-profile-edit <profile-id>
command will be added. With due consideration for authentication and authorisation, the behaviour of this command will be:
-
Retrieve the current profile content (in the existing key-value format used for file-based profiles, rather than LDIF, JSON or other.)
-
Save the content to a temporary file.
-
Invoke an editor on the file. Respect the <code>EDITOR</code> environment if set, otherwise invoke
vi(1)
. The user makes changes, saves the file and quits the editor. -
If changes were made to the profile, store the updated profile in the database (the change will be automatically replicated to clones). If no changes were made, report that no changes to the profile were detected.
-
Remove the temporary file.
Other useful operations that could be implement as subcommands of pki ca-profile
include:
-
Showing a diff between a profile and the system/default version of that profile (if it exists).
-
Creating a copy of a profile, under a different name (most likely for subsequent editing.)
Updates to profiles via the CLI tool shall not require a restart of the pki-tomcatd
service.
Existing access controls shall remain. That is:
-
Update of an existing profile - agent disables the profile; admin then is allowed to update; agent reviews the profile and enables it.
-
Adding a new profile - admin creates the profile; agent approves it.
The implementation will be done in stages. Additional requirements or changes discovered during the implementation process will be detailed for each stage of the implementation. Patches will roughly correspond with each stage.
Update ProfileSubsystem
to use the LDAP database instead of files.
The LDAPConfigStore
class was added. It implements IConfigStore
but since none of the profile code uses the backup
feature, the commit
method ignores this argument. The method is documented to explain this, and recommends an approach to implement backup should it be needed in the future.
Implement script(s) for importing file-based profiles into the database.
The import procedure is implemented as part of the CAInstallerServer
process.
The main issue encountered was that the ProfileSubsystem
, after being modified to talk to the database, cannot start up until the database connection is configured. This was resolved by adding support for disabling dynamic subsystems in CS.cfg
, as well as the methods CMS.enableSubsystem(String id)
and CMS.disableSubsystem(String id)
. The profile subsystem is initially disable, but is enabled during the spawn process as soon as database configuration is completed.
Add the --raw
flag to existing CLI commands for working with the "raw" profile config format, and implement the pki ca-profile-edit
CLI command.
The ProfileResource
REST API required a few new methods for working with the "raw" (i.e. ConfigStore
) profile format, as a byte[]
, instead of the default XML/JSON profile transport format. The new methods are:
-
createProfileRaw
-
modifyProfileRaw
-
retrieveProfileRaw
The ProfileClient
API works with Properties
objects instead of the raw byte[]
objects, to make things more "developer-friendly". A caveat of this decision is that Properties
keys are not ordered. Users of the ProfileClient
API must therefore sort keys themselves if they wish to present sorted profile config properties to end users.
Implement profile change replication monitoring and refresh mechanism.
Some methods of the ProfileSubsystem
were made synchronized
in order to safely handle updates from the LDAP persistent search thread that monitors for updates to profiles.
The LDAP persistent search implementation is entirely encapsulated within ProfileSubsystem
and cannot be access or controlled by the user.
One LDAP connection is held at all times by the persistent search thread.
Implement upgrade scripts for initial import of file-based profiles into the database (using the script(s) from earlier).
This is expected to use the upcoming database upgrade framework.
The ProfileSubsystem
will need to be initialised such that it has read/write access to the database.
Parts of CS.cfg
and the registry will become obsolete, and can be removed.
There remains the possibility that users will decide whether to use LDAP profiles or file-based profiles. If this is allowed, corresponding configuration options and (most likely) pkispawn
options will need to be added.
-
10.3 → 10.3: This proposal does not present any new concerns for cloning a 10.3 database using Dogtag 10.3.
-
10.3 → 10.2: Cloning a 10.3 database using Dogtag 10.2 will be prohibited.
-
10.2 → 10.3: Cloning a 10.2 database with Dogtag 10.3 will be permitted. The 10.3 installation will include LDAP-based profiles. Modifying (file-based) profiles on the 10.2 installation will have no effect on the 10.3 installation. This is a continuation of the present behaviour with file-based profiles. Upgrading the 10.2 installation to 10.3 at a later time may result in conflicts. A strategy for dealing with these conflicts needs to be determined.
(edewata) I’m not sure if we should support 10.2 → 10.3 cloning. When we release 10.3 the 10.2 will still be fairly new so it might be reasonable to require all clones to be upgraded. It will reduce the amount of testing requirement too.
CS.cfg
may require updating, as explained above.
Users should be alerted (via release notes) of this feature, and instructed to disable any custom mechanisms they may have in place to replicate profile changes between clones.
The 10.3 migration process must move all profiles into LDAP. File-based profiles will be left on the filesystem for the time being, but will no longer be used.
A database attribute will record whether a profile was user-defined or user-modified, for use by update scripts.
Because behavioral changes to default profiles are rare, this design proposal does not specify a mechanism for handling them. Such changes should be managed on a case-by-case basis by optional update scripts (i.e., not run automatically, but at the administrator’s discretion). Accompanying release notes should explain the behavoiural changes and detail the process for applying the changes.
ORIGINAL DESIGN DATE: June 20, 2014
One of the two initially-proposed solutions was a hybrid LDAP/files solution, where system profiles continued to be stored on the filesystem, but modifications could be stored in LDAP, and all custom profiles would be stored in LDAP:
-
Profile creation will store the new profile in LDAP, so that it will be replicated.
-
Modification of a file-based profile will result in the modified profile being stored in LDAP, so that it will be replicated. Consequently, the LDAP profile storage must take precedence over file-based profile storage in the profile lookup process.
-
Because LDAP and file-based versions of a single profile may now exist at the same time (the LDAP version being the active version), the behaviour of the delete profile operation needs to be clarified. Because System profiles proposes using the shared system profiles (which an instance will not be able to delete), I propose that Dogtag prohibit the deletion of profiles that have a file-based version (whether or not there is also an LDAP version).
-
If there is a use case for restoring a profile to the default version distributed or installed by Dogtag (where it exists), a new restore profile operation can be implemented. This operation would remove the (modified) profile from the LDAP directory. The file-based version will then become the active version. Attempting to restore a profile that exists only in LDAP would be an error.
The main motivation for this proposed solution was to simplify application of updates to default profiles:
-
When upgrading to LDAP-based profiles, upgrade scripts must detect added or modified profiles and move these into the LDAP profile storage. Added profiles will then be removed from the CA subsystem profiles directory, and modified profiles will be restored to a pristine state, which will ensure:
-
updates to default profiles can always be written to the corresponding file-based profiles without conflict;
-
a smooth changeover to a System profiles directory will be possible, if this proposal is implemented.
-
alee had reservations:
-
I understand why you have profiles in both LDAP and file format. However, I think this makes things complicated. My preference would be to have all new systems maintain their profiles solely in LDAP, rather than some admixture.
-
There is a precedent for moving data that was formerly in files to LDAP - and that was the data in the security domain. Originally, this data was in files. At some point, we changed the servlets that update the security domain to use LDAP instead, and used a parameter in
CS.cfg
to determine whether the data was in LDAP or files.
edewata proposed a variation where only custom profiles would be stored in LDAP, and default profiles would continue to be managed on the filesystem, as they currently are.
-
I think all system/default profiles should remain file-based and all custom profiles should be LDAP-based. It will make a clean separation: system profiles are owned by us (Dogtag developers), custom profiles are owned by the admin.
-
The system profiles will be read-only. This way we will be able to update the system profiles without writing any upgrade scripts because the files will be updated automatically by RPM. Just one requirement, all server instances must be upgraded to the same version.
-
If the admin wants to change a system profile, they can clone it into a custom profile and make the changes there. The custom profiles cannot have the same names as the system profiles, so there’s won’t be any conflict/confusion, and no need to support a "restore" command. In general we won’t need to write upgrade scripts for custom profiles except if we change the LDAP schema.
One significant point in favour of edewata's variation is that administrators can continue to manage profiles in the way they are used to, i.e. editing them directly. The pki ca-profile-edit
CLI is deemed to be a sufficient mitigation.
Due to the rejection of automatic updates to default profiles (see below), which was the primary motivation for the files/LDAP hybrid solution, and in consideration of the increased complexity, the hybrid solution was rejected.
The original proposal for LDAP-only profiles was to automatically effect behavioural changes to default profiles as part of the upgrade process:
-
There is currently a 10.3 ticket to create a database upgrade framework. Once this framework is in place, it can be used to perform a migration from files to LDAP, as well as modify default profiles when the default profile is being used.
This was rejected, although tools will still be provided for an administrator to perform the update at their discretion. alee explains:
-
There is another problem, and that is that it is not clear that we want updates to the default profiles to be propagated to existing instances. I have looked at the profiles and there have been only a handful of changes over the last 7 years. Those changes include things like updating the default signing algorithms or the default validity. More likely than not, admins would prefer that we not change the behavior of profiles in existing instances underneath them.
-
The changes that I have found are all behavioral - and therefore things that admin can opt out of — or would prefer to do on their own schedule. There have been no structural changes.
-
If there are structural changes, then we need to (and can) provide an upgrade script which would run with the automatic upgrade. An example of this would be a schema upgrade as we sort out how to represent profiles in LDAP.
edewata proposed a fine-grained storage of profile data, instead of simply storing the current profile data as a single bytestring (in the same way that all the profile data is currently stored in a single file):
-
I suppose we want to have something that resembles the actual Profile data structure (see
ProfileData
Java class). There should be an LDAP attribute for each single-valued Java attribute (e.g. name, description, enabled, visible). This way the profile is more manageable and can be queried based on these attributes. For collection attributes (e.g. inputs, outputs, policySets) we can use child LDAP entries to represent them. -
About the REST interface & CLI, since this will be the primary way to edit profiles, we might want to have more granular commands to modify parts of the profile. Right now with ca-profile-mod command you need to send the entire profile in a file. It would be nice to be able to specify some parameters to change certain attributes only, or use separate commands to manage the inputs/outputs.
-
We’ll also need an interface to find existing cert records that use a certain profile and bulk modify them to use a different profile. This will be useful when you create a clone to change the system profile.
There are obvious benefits to this proposal but it is more work (the existing machinery for reading and modifying file-based profiles would no longer be useful for LDAP profiles), and not necessary to maintain the current behaviour and meet the basic goals concerning replication. It is therefore deferred.
edewata proposed a mechanism whereby profiles can inherit from other profiles:
-
Basically each LDAP profile will have an optional parent. The parent can be the file-based system/default profile, or another LDAP profile. A sub-profile will inherit all attributes, except when it’s explicitly declared in the sub-profile. This mechanism allows us to create just a proxy/alias, a full clone, or anything in between. For example, a proxy profile might only have a few attributes:
dn: cn=caAdminCert,ou=Profiles,ou=CA,{suffix} objectClass: certProfile cn: caAdminCert parent: defaultAdminCert visible: true
This proposal was deemed to be out of scope with respect to current requirements but fundamentally compatible with this proposal, and was therefore deferred.
LDAP-based profiles mean that there are more moving parts and different kinds of errors are possible. This section is designed to help you troubleshoot issues that may be related to the LDAP profiles subsystem.
By default, PKI is configured to use the ProfileSubsystem
class which uses file-backed profiles.
When deployed as part of FreeIPA, since FreeIPA 4.2 (RHEL 7.2) the LDAPProfileSubsystem
class is used. The FreeIPA 4.2 upgrade process switches deployed CA instances over the LDAPProfileSubsystem
and imports profiles into the database. There were a couple of bugs in the upgrade/migration procedure that could result in missing profiles:
Execute:
$ grep subsystem.1.class /var/lib/pki/pki-tomcat/ca/conf/CS.cfg
The output indicates the subsystem in use, e.g.:
subsystem.1.class=com.netscape.cmscore.profile.LDAPProfileSubsystem
To see which profiles are present in the database, execute:
$ ldapsearch -D "cn=Directory Manager" -w $DM_PASSWORD \ -b "ou=certificateProfiles,ou=ca,$BASEDN" -s one cn
(Replace $DM_PASSWORD
and $BASEDN
with the appropriate values. When deployed with FreeIPA, the base DN is o=ipaca
.)
If the LDAPProfileSubsystem
is being used, the output should show ~60 profiles.
If the output is:
# extended LDIF # # LDAPv3 # base <ou=certificateProfiles,ou=ca,o=ipaca> with scope oneLevel # filter: (objectclass=*) # requesting: cn # # search result search: 2 result: 32 No such object matchedDN: ou=ca,o=ipaca # numResponses: 1
Then the ou=certificateProfiles,…
container is missing.
If the output is:
# extended LDIF # # LDAPv3 # base <ou=certificateProfiles,ou=ca,o=ipaca> with scope oneLevel # filter: (objectclass=*) # requesting: cn # # search result search: 2 result: 0 Success # numResponses: 1
Then there are no profile entries present at the correct location in the database.
Prior to the fix for bz1300252 being released, you can trigger the migration of profiles into LDAP to be attempted again.
Edit /var/lib/pki/pki-tomcat/ca/conf/CS.cfg
and replace the line:
subsystem.1.class=com.netscape.cmscore.profile.LDAPProfileSubsystem
with:
subsystem.1.class=com.netscape.cmscore.profile.ProfileSubsystem
Then execute ipa-server-upgrade. The upgrade program should observe that LDAP-based profiles are not enabled, re-enable the LDAPProfileSubsystem
and migrate all file-based profiles into the database.
Tip
|
To find a page in the Wiki, enter the keywords in search field, press Enter, then click Wikis. |