Wednesday Sep 14, 2011

Global Policy Attachments - Inheritance rules - Part#2 - 11g

In this blog post i briefly mentioned about GPA vs. LPA and when to use GPA. As I mentioned the key difference between GPA and LPA is around granularity (or scope of the policy attachment). So what are inheritance rules - if we define GPA that applies to say "all domains" vs. a GPA for "domain1" vs. a GPA for "app1", etc then the inheritance rules determine which policy get's enforced for a particular web service.

Broadly speaking there are two types of inheritance rules:

  1. Overriding rule
  2. Additive rule.

So here is a simple scenario.

Scenario#1: We have a deployment with two weblogic domains (Domain#1, Domain#2). We want to secure all web services (SOA, ADF, etc) in this deployment with wss11_username_token_with_message_protection_service_policy as shown in the figure below.

GPA all domains

Click here for a larger image.

Here are the WLST commands to setup GPA for the above scenario:


'ws-service', 'Domain("*")')

$>setPolicySetDescription('Default policies for web 

services in any domain')


See here for detailed description of these commands.

GPA Overriding rules

Scenario#2: If we now want to consider a scenario where in for a particular app (let's say "GeneralLedger" needs to be secured with oracle/wss11_x509_token_with_message_protection_service_policy) - then we need to define a new GPA. This is shown in figure below:

overriding rules

Click here for a larger image.

Here are the set of commands to define the GPA for GeneralLedger app to be secured with oracle/wss11_x509_token_with_message_protection_service_policy.


'ws-service', 'Application("GeneralLedger")')
$>setPolicySetDescription('Policies for web services in General ledger app')

In the scenario above - for Web Services in GeneralLedger the policy oracle/wss11_x509_token_with_message_protection_service_policy will be applied. Thus the application specific GPA overrides the deployment wide GPA. So we have the first rule which is basically that the more specific GPA overrides the more generic GPA.

GPA Additive rules

Scenario#3: Now let's consider a scenario where an application with a single WS (say "Reliable & Secure WS" - for lack of a better name!) want's security and WS-RM. Also we want the security to be the same as the deployment wide posture i.e. the app needs to be secured with oracle/wss11_username_token_with_message_protection_service_policy. In this scenario - all you need to do is attach oracle/ws_reliable_messaging_policy via LPA to the "Reliable & Secure WS". In this case OWSM recognizes that the "category" of the policy defined at the GPA level is "security" and specifically it is "authentication" and "message protection" subcategories under security and this in this case adds the policies such that the policies applied for "Reliable & Secure WS" is both oracle/wss11_username_token_with_message_protection_service_policy and oracle/ws_reliable_messaging_policy.

This is depicted in the figure below:

additive inheritance rules

Click here for a larger image.

Since the inheritance rules change based on the "category" of the policy and it may not always be clear as to which policy is being applied to a particular web service - OWSM provides what I call the "Effective Policy View" i.e. the set of policies that will be applied to a Web Service after applying of the inheritance rules I just described above. You can view the "effective policies" either in EM or via WLST. See this section in the documentation for a description around "Effective Policies".

Note: GPA is currently as of 11gR1 PS4 not supported for WLS WS and OSB.

Policy Attachment - GPA vs LPA - Best Practice - Part#1 - 11g

In a previous post I briefly mentioned about the fact that in OWSM 11g - we support Global Policy Attachments (GPA) and Local Policy Attachments (LPA) or sometimes also referred to as Direct Policy Attachments (DPA). In this post - I will provide some best practice guidelines on when to use GPA vs. LPA.

For those who are not familiar - in the case of LPA you attach it to a WSDL Port - so the policy applies to WSDL Port. GPA allows you to attach a policy to more coarse grained entities - ex: it can be to a Weblogic Domain - in which case the policy applies to all WSDL Ports running in that weblogic domain. Or the policy can be attached to an application - in which case the policy applies to all WSDL Ports that are part of that application (ex: ear). etc...

So the first main difference b/w GPA and LPA is the granularity. So when would one use GPA? Basically two scenarios:

a) You want a "Secure by Default" story

b) Ease of management for large deployments.

Secure by Default

So what do I mean by "Secure by Default"? Let's say you develop a web service in your favorite IDE. You can secure it at Design Time. However let's say you decide not to secure it at design time. Now the app is deployed - this app is not secure unless somebody goes and secures it post deployment in EM or via WLST. If there are no strict controls and processes in place - that ensures that app is secured before it is made available to the outside world - then you have a potential for vulnerability in that the app is running unsecured.

Improved manageability

Let's say you have a large deployment - 100's of web services - lots of WLS servers, multiple weblogic domains, etc. In this case going and securing each of the 100's of web services can be tedious, time-consuming, and error prone - especially if you want to ensure there are no web services that are unsecured. In such cases, if you have standardized on the security posture of your web services - GPA can be a life saver!.

You can define a GPA that says "all web services (WSDL Port) in all domains" in your deployment use let's say wss11_username_token_with_message_protection_service_policy!

Now whenever a new app is deployed to one of the domains in your deployment - it will "automatically inherit" the oracle/wss11_username_token_with_message_protection_service_policy and you don't have to go into individual web service and secure it. Let's say a year from now - you decide to change the default security posture to say all oracle/wss11_x509_token_with_message_protection_service_policy - then all you need to do is change it your GPA definition - in one place and all the web services in your deployment will be secured with this new policy!

When should you not use GPA?

  1. if you have not standardized on a security postured for your web services - then GPA is not very useful!
  2. Using GPA for role based authorization policies is not very useful. Typically different web services will be accessible by different roles. (Note: It would be ok to use GPA for permission based authorization policies).
  3. If your policy has app specific aspects - then GPA is not appropriate.
    • If you want specific parts of the message to be signed or encrypted as discussed in this post
  4. If a policy has expectations around how code is written by developers then GPA is not appropriate.
    • Ex: Using GPA for MTOM or WS-RM may not be appropriate. Not all web services support attachments - especially MTOM attachments. Also in many cases WS-RM or MTOM may require some coding considerations by developers.
Here are some pointers to the OWSM documentation on GPA and LPA.



Note: The documentation uses the terminology PolicySet for GPA. PolicySet is the underlying implementation model for supporting Global Policy Attachments!

GPA is extremely powerful - but you need to really understand the pros & cons before you decide to use this feature. In future posts - I will discuss:

a) some of the inheritance rules, etc associated with GPA and also what happens when you have GPA and LPA in a deployment or what happens when you want to mix policies from different categories - ex: Security and WS-RM.

b) When to combine different granualrities

c) Life-cycle aspects of GPA and how it differs from LPA.

Tuesday Sep 06, 2011

Attaching OWSM Policies - Best Practices - 11g

As mentioned in previous post - there are three ways to attach Policies in OWSM 11g.

a) at Design Time (DT) in an IDE - ex: JDeveloper

b) Post Deployment - using WLST or FMWCTL

One of the questions that pop up from time to time is around when to use what - in this post I provide some guidelines that can help in deciding which is the best methodology.

Before I actually layout some guidelines - we need to discuss a related feature supported in OWSM 11g - this relates to the fact that attaching a policy at DT does not prevent you from changing it post deployment. So (a) and (b) are not necessarily mutually exclusive. The reason OWSM provides this flexibility is in many organizations - administrators (either app administrators or security administrators) decide on the security posture and hence the ability to change the security posture by changing the policy of a Web Service or Web Service Client is an important consideration [1].

So if your organization is more attuned to the process of standardizing security upfront and communicating it to your developers then developers can secure it at DT, test it early to ensure all aspects are working before the app gets deployed to a testing, staging, production environments. In general I would recommend some level of testing with security enabled in your DT environment. (a) is helpful in those scenarios.

(b) is relevant in two scenarios:

i) you have unsecure apps that need to be secured by the administrators post deployment.

ii) you have secure apps but they don't adhere ot the security guidelines standardized by the organization and hence you need to change the policy attached to the WS/WS Client.

Using WLST vs. FMWCTL:

WLST is more suited if you would like to script things out (in addition some administrators like command line tooling while others prefer a Web based user interface).

FMWCTL is more suited for people who prefer a Web based user experience or for non-scripting scenarios.

In future blog posts I will discuss guidelines around when to use Direct Policy Attachments (or Local Policy Attachments) vs. Global Policy Attachments (GPA).


[1] In - WLS JAX-WS Client support only a programmatic model of attaching the policy to the client app. In this scenario - one cannot change the policy post deployment.


In this blog I will discuss mainly features supported by Oracle Web Service Manager (OWSM).


« April 2014