Red Hat OpenShift OpenShift Commons Briefings, 26 Jan 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Open Policy Agent Gatekeeper: Centralized Policy and Governance - Jaya Ramanathan (Red Hat)

Description

Open Policy Agent (OPA)Gatekeeper: Centralized Policy and Governance

Security and Governace Strategy and Architecture
Advanced Cluster Management for Kubernetes

Guest Speakers:
Jaya Ramanathan (Red Hat)
Yu Cao (Red)
Moderator: Kirsten Newcomer (Red Hat)

A

Hi everyone. um Thank you for this opportunity. um I am nothing. I'm the chief security and governance architect for our advanced cluster management offering- and I have with me youtau. He is the squad lead for our security and governance capabilities within both of us work on this project called advanced cluster management for kubernetes within red hat, and we also have an upstream community called open cluster management.

A

So we are here to talk about just some underlying principles and overall architecture of our security and governance capability, and then we will delve deep into oppa and gatekeeper and how that integrates with acm. Okay, um this chart basically talks about the various security and compliance challenges that customers face when they adopt cloud uh specifically.

A

The key thing I wanted to highlight here is uh customers generally have to meet various enterprise standards, as well as industry standards, for example, if they are in regulated industries like financial healthcare, etc, they have to meet standards like pci, hipaa, etc, and they also want to adopt open technologies, and they typically have to deal with more than one cloud provider, and they also need to go through multiple audits.

A

So really what they are looking for is the ability to maintain security and audit readiness on a continuous basis, and they want to do for open, hybrid cloud.

A

um So our approach um to achieve that goal of continuous security and audit readiness is policy-based governance. So what do we mean by that?

A

Typically, if you take a particular um cloud environment, say red hat, openshift uh cloud platform, and you want to secure it um you, the customer, would enable various security capabilities that openshift provides out of the box, and they may also want to use some capabilities provided by third party and some homegrown capabilities as well and and what they would like to do is to ensure that all these security controls are configured to industry, best practices or their enterprise standards, and these standards could include standards from nist lightness, csf nest, 853 and also, like I mentioned industry segment standards like pci, hipaa, etc.

A

So the subject matter expert for a particular control uh knows very well how to get that control set up. But if you're talking about sre or a sec ops person who is responsible for ensuring that the configuration is set up properly, they may not be an expert on all the aspects of security.

A

So how do we make this easy for them? So the way we make this easy for them is by ensuring that these best practices are represented as policies that result in a desired configuration state.

A

So that's the overall approach and that's what we refer to as policy-based governance and in order to do this, what we have built out is an extensible policy framework that can be applied across the entire stack.

A

It allows you to incorporate multiple policy languages easily, and so that's what we'll talk about in detail, the gatekeeper which is one of the policy languages and we'll also, you will also show a demo on how that can be done and we provide a lot of out of the box content, but we want to make sure those that content is customizable and we need the ability to easily integrate third party controls, um etc.

A

So that's kind of the overall approach of policy based governance, so I want to introduce some terminology here just to level set uh what you're talking about here. So if you think about policy based governance right, so there is a policy authoring point where the policy is specified.

A

Now this authoring point could be either the console or it could be a cli or it could be githubs. Our preferred methodology is githubs, the reason being that allows you to manage policies. Just like you will manage source code, so you define all the policies in git and then within our acm product you can define a subscription to pull those policies and deploy them onto the managed clusters.

A

So that is our policy management point. So this distributes the policy. It consolidates policy violations across a fleet of clusters. It also allows you to enter integrate with enterprise tools from a central point, because that is how customers view the cloud as just an extension of the of their existing id infrastructure. So they want uh the management to integrate with their incident management systems, uh security, operation center, their enterprise, grc tools, etc.

A

And then, when the policy gets distributed to the managed cluster on the manage cluster, you have policy enforcement points and these these consume the policy and they return the violations. So an example of enforcement point would be the kubernetes admission web hooks and also the runtime controllers. Communities controllers themselves, as well as several controllers that we provide um the configuration policy controller and the certificate management controller, as well as any operator that runs on openshift.

A

Manage cluster can also serve as a policy enforcement point and one one of the examples of the operator is the compliance operator that allows you to implement security profiles on your openshift clusters.

A

So and then the policy enforcement point can optionally uh invoke a policy decision point to check whether a specified policy matches how control is configured now. In some cases, the policy enforcement point may itself do this check, but in other cases um you could invoke a pdp and uh the gatekeeper oppa integration is one such example and cubano is another example.

A

So, based on this terminology, um here is the overall architecture. So, as you can see here on the management hub, um this is where the policies are defined and, as I mentioned, it can come through from git or it could be through ui or cli, and then from here it gets distributed to the managed clusters. And then here you have the various policy consumers. These are essentially the policy enforcement points that I talked about and then the these results are then collated back and, uh and so you have everything available uh in a central manner.

A

uh There are a few roadmap items that we are highlighted here, where we want the policy framework uh to associate ansible playbooks to trigger automation, automated playbooks, for policy violations, and we also want to integrate the policy violations and generate alerts to the observability component, which can then integrate with enterprise incident management tools, and we also want to route audit logs to sims for the security operations center.

A

These are all the orange items in this chart are things that are roadmap items so now I want to drill down into details of how we have integrated with gatekeeper and oppa. So let me start with the quick overview of gatekeeper and oppa, so gatekeeper allows you to evaluate compliance of kubernetes resources to specified policies and what it leverages under the covers is oppa and oppa is the policy engine that it uses now.

A

Oppa itself accepts policies in a language called rego, so think of gatekeeper as providing the bridge from oppa to kubernetes to the kubernetes world. So it allows you to specify the policies as kubernetes constraints, which can which can in turn contain the rego constructs and gatekeeper, can then consume those and um because it registers itself as a policy enforcement uh point through the web hook, and um so it's now able to pass on those regular constructs to oppa and use oppa to do the actual uh checks of whether the policy is complied to.

A

So two scenarios to consider here one is the admission scenario, so this is uh executed whenever a cumulative resource is created or updated, and um so so think of it. As once, the gatekeeper policy is in place, then any uh new resources that get created will be governed by this admission control, but what if some resources already exist before you put that policy in place? So that is the second scenario, which is the audit scenario. So the audit scenario allows you to periodically evaluate existing resources against policies to detect any non-compliances.

A

So in order to complete the picture, um there were a couple of things we had to do in in terms of how we integrated uh gatekeeper and opa into acm. First of all, we deliver. uh We are delivering going to deliver a operator that can deploy gatekeeper and oppa, and uh so this is this. Work is actually progressing in the upstream uh community and what we have done is we have for the community and made it part of our open cluster management uh downstream.

A

um So we will deliver this as part of the acm product. So one of the cool things about the acm product is our built-in policy controller. Config policy controller can pretty much distribute any kubernetes crs. Now the deployment of operators itself is represented as a cr, so by delivering gatekeeper opa as an operator now you can just define a acm policy that ensures that this operator is deployed on the managed clusters, so that so that's uh so we are going to provide that.

A

In addition, uh we also you can also use acm to uh define policies to on what gatekeeper itself is enforcing right. So, though, the constraints that I talked about so those can be also authored within rackham and uh propagated to the managed clusters. And then, as I mentioned, there are two scenarios admission and audit scenarios. So in both cases we want to be able to detect policy violations.

A

So in the case of the admissions scenario we just, we have a rackham config policy that can process the events that are generated by the gatekeeper webhook and in the case of audit scenario, we look for the status field in the gatekeeper constraints. So so we are able to pull both this information back, as I showed in the architecture into the central acm hub and um and display the results there.

A

So this uh architecture diagram pictorially, depicts um what I talked about. So you have the rackham hub and then, which is distributing policies to your managed cluster and within the manage cluster you have the config policy controller and because we have this built-in config policy controller, the whole integration with gatekeeper opa is happening just through the authoring of these policies. So you really don't have to write any code.

A

So so we have a policy that we have authored to deploy the gatekeeper operator, so that operator is then watches over this gatekeeper, cr and and then it registers this webhook that monitors um the audit uh the admission scenarios and then uh you will also have a policy to deploy the gatekeeper constraints.

A

So and then we have a policy, as I mentioned, to detect the violations, two policies, one for audit and one for the admission scenarios.

A

So you is going to give us a demo on on these capabilities, um but before I hand it over to you, I also want to highlight um that we have also. We also have integration in place with acm for the openshift compliance operator, uh which is another an example of an operator that can be deployed on the managed cluster and through a rackham policy.

A

We can make sure this operator is deployed, and we also have another policy that allows you to specify a profile security profile that directs the compliance operator to perform the compliance checks and return results back to us. So we'll show a quick demo of that as well.

A

That kind of is the quick short overview um and I'm I have in this uh chart a link to our open cluster management community, and you can connect with us there and and ask any questions etc before I turn it over to you. I also want to give you a quick demo of our open cluster management community.

A

So, as I mentioned when I uh provided the overview of policy based governance, typically, you have various controls in place and the controls could be for security could be for resiliency could be for software engineering, and you want to ensure that all these controls are configured properly to industry, standards and enterprise standards, and in order to do that, our goal is to codify those standards as policies, and so it's very clear that red hat on our own cannot produce all the policies needed by all our customers.

A

So that's why we have created this policy collection repo, so it's a spot for us to collaborate to build out these policies. So if you look at this repo, we have two folders. We have a community folder and we have a stable folder.

A

So the stable folder includes all the policies that are delivered as part of the acm product. They are fully supported by red hat and these policies are organized in terms of the nist 853 standard.

A

So you have various uh controls within this security controls within this catalog and um and then for each of the controls you can see, we have defined policies so, um and you is going to show us a demo of the compliance operator policy that can deploy the compliance operator and also this um essentially security profile policy that allows you to scan the managed cluster for this particular security profile and then for gatekeeper and opa. We have policies as well, and then there is the other folder here, which is the community folder.

A

So the community folder is where we invite contributions from the community, and these policies are also organized in terms of the nest. 853 standard and the gatekeeper policies currently are in the community, and the reason is because we are currently working on an acm release where we are productizing this capability, so they write down sit in the community, so you have a gatekeeper operator policy that allows you to deploy the gatekeeper operator and then we also have other policies for gatekeeper, contributed here as well.

A

um So so you can see here, we also have third-party contributing policies here. So you will see here a policy from cystic. uh They have contributed a policy for their operators. So uh the reason I wanted to highlight this is to invite contributions from the community, so we are also working with various customers to have them contribute here as well.

B

If people, I think the ocm team group is meeting on a semi-regular basis now have they started the community meetings.

A

Yes, we have started the community. The open cluster management um has a community meeting that happens weekly and uh definitely you know, invite uh participation in that as well. So I can uh include that link as well into the uh chart.

B

Okay, that would be great um and there's there's one quick question here before we go to the demo. um Tobias is asking: if, if the gatekeeper community would announce mutating admission controller, will you integrate such an option into rackham.

A

Yes, uh so we are active participants of the gatekeeper community. um You who is here on the call with me, along with a couple of others from red hat, are active contributors there as well, so we are watching the space there and our goal is to keep up with any enhancements that are made in gatekeeper and associate policies and track them to adopt those. Yes, that circle.

B

I am so grateful that you finally said that acronym out loud and it's rackham, um because I have that's- that's a wonderful way to pronounce it, so I'm going to use that going for it. So thank.

A

You uh that's a good point. Let me clarify rackham and acm stand for the same thing: acm is advanced. Cluster management and rakham is red, hat advanced cluster management, the same product offering and our upstream community is the open cluster management community, which is this one.

B

And ginny just shared a link in the chat for the open cluster management community. There's a projects page in the repo that you can find the agenda and things like that for the upcoming meetings.

A

um I'm going to stop sharing now and turn it over to you. Perfect yeah, welcome, further questions as well. Thank you.

C

All right, uh let me share my screen. First.

C

Okay, all right um so um hi everybody, um my name is yuzhou, so I'm the uh dev lee for welcome security and compliance team. So today I'm going to do a quick demo for how you can use rackham policy to install gatekeeper and then apply a gatekeeper constraint on the manage cluster and then report any violations it detects. So, first of all, um let me recap a little bit about the the governance and risk architecture in uh rackham products. So if you go to our product documentation, there's an overall architecture diagram here.

C

So if you click on it, you'll see a large diagram, so the the governance policy framework we put in place. It really consists consists of multiple piece right. So there's a one piece on the top and there's a few piece on the manage cluster so on the left so on hub. So we have policy prop data which responsible for propagating the policy from hub to manage cluster, so how it how it does is so when you create a policy, you need to also create a placement binding and placement rule.

C

So if, if you look at a sample here, so we have a policy crd, which is the crd that we use to embed your embed, your the real policy. So if you look at the crd definition here, so we have a field called policy templates, so inside templates you define the actual policy. You want to apply on your manage cluster, so you can you can you can create a policy with multiple policy templates using different policy language and then, once this is created, you create also create a placement rule.

C

So since recom is managing the purp. Is a control plane that managing multiple clusters right? So um so, when you import a cluster, you can label your cluster, so the placement simply here is you can select a class cluster either you using the the label selector, so we call custom selected. But what really does is it's using labels on the cluster so that you can?

C

You can select a subset of cluster of a device, for example, in this case it's all your dev clusters and then create a placement binding that find the policy and placement together basically tell our policy propagator.

C

This policy needs to be applied to these clusters, so this is how we propagate policy to manage cluster. So the reason I explained this is so. This is the framework that to distribute policies and the actual uh actual the actual policy that is being executed or evaluated on the manage cluster is we call it policy controllers or policy consumers on the manage clusters, so the framework propagates, the policy to manage cluster, distribute them to manage cluster and then hand it over to the actual policy consumer installed. On your manage cluster, for example, the gate gatekeeper.

C

Policy, consumer or compliance operator or the out-of-box policy controller that we ship reconship, for example the configuration policy controller. So once it is, it reached the magic cluster. The the actual policy consumer on the magic cluster will will apply, evaluates the the policy past uh passed in and then um return and then generate results and the policy framework will retrieve those results and return it to the hub. So that you can uh you can view the results uh through single control plane using recon console.

C

So that's the governance uh framework policy framework and we put in place for blackcom um so in terms of gatekeeper policy. So from this big picture, so gatekeeper uh controller is one of one of the policy consumer from recon policy policy framework perspective.

C

um Now so now, let's take a look at the actual gatekeeper policy, so we've talked about that. We can use that raccoon policy to install gatekeeper and we can also use a record policy to apply gatekeeper constraints and then collect the results back to the hub. So this is done through defining record policy. So, as jr mentioned in this policy collection repo, you can find you can find the gatekeeper operator policy and also some a few other uh gatekeeper uh policy that actually deliver applying certain constraints on your clusters.

C

So let me take a sample here. Let me go through in details. uh For example, the first uh the gatekeeper operator, gatekeeper operator is, is an f as an effort that we will be actually involved in the upstream community, so the geek. What the keeper operator does is it packs it packages the gatekeeper as a openshift operator, the operator responsible for the life cycle of gatekeeper, so it can install gatekeeper, it can delete gatekeeper. You know, of course we are. We will.

C

We are also working on upgrading those kind of stuff, so it controls the whole lifecycle gatekeeper. It simply makes installing gatekeeper easier, with gatekeeper being delivered as a gatekeeper operator in orm or in ocp catalog. You can use a raccoon policy to to insta to to install gatekeeper across all your manage cluster. So what you need to do here is you define a policy uh crd policy cr here and then it includes multiple templates. So uh just very similar to how uh how you install uh an operator from om, so you create a namespace.

C

We defined a catalog source. Currently the gatekeeper operator is not published yet so we have to define a custom catalog source, um and then you create a operator group to tell operator which namespace to operate on and then the subscription to subscribe, the the catalog um to pull down the gatekeeper uh images and and so and then um and then you create this gate, gatekeeper cr. This keeps it. Keepers is part of gatekeeper operator, so it tells you how to configure gatekeeper.

C

So if you create this cr8, the gatekeeper operator will install the gatekeeper based on the spec. The configuration provide. If you delete this gamepcr, the gatekeeper operator will delete the gatekeeper instance on your cluster. So here, as you can see, we are pulling an image from from upstream gatekeeper as this is, uh and then we enable the uh admission events flag so that we can. We can so that when there's a violation january for uh ford for any new, when there's a violation generated generally for any new, is the new resource cr created unmatched cluster?

C

That being that was, that's is blocked by the skate keeper. uh It will generate equipment, kubernetes events, so this is for the purpose. This is for the purpose for raccoon to be able to collect the result for a diminishing admission um scenario. So then, so, once this this, this policy is created, you will and um created. uh Then uh it will based on the cluster, is selected here right, so it will install the the gatekeeper operator on manage cluster so for this is the raccoon console. So I have a gatekeeper operator policy installed here.

C

So, as you can see, this policy currently is being applied to this. This managed cluster. um So if you go to this manage cluster here, it was so this is currently being this part, is being forced and is compliant. So this means the operator has been installed on your manage cluster. So if you go to the ocp console here on this manage cluster, this is not a managed cluster. Let me open.

C

This is the match cluster and then, if you go to operator here, go to install operator, you'll see gatekeeper operator here, so it is so very simple. So the you just need to use this operator. um You just need to use raccoon to install this operator for you and then the gatekeeper is installed on manage cluster.

C

So now, let's take a look at how you can apply gatekeeper constraints and then report report violation back. So installing gatekeeper is just the first step right. What was really what really has value is you want to apply gatekeeper, constraint and constraint, template and then collect results back so um here we I create a sample.

C

I create, I create a gatekeeper sample policy here, so I think it's easier to look at in here in vs code, so gatekeeper. So this is from the gatekeeper. This is from the policy collection input um you can find it. You can find it in collection. Repo, so same here we are using the same policy cr to distribute the gatekeeper constraint, constraint, template as you can see. We have three templates here.

C

If we look at the details right this, this is actually leveraging the configuration policy, which is the out of box policy consumership uh uh in recon. So what it does is um what it does is. It can create update, delete resources um on the cluster uh based on the template object, the definition you provide. So if you look at the sr cr fields here, so you create a configuration policy, it has object templates inside the spec. So in this object templates anything defined in this object.

C

Template template field is the actual cr you want to create in this example. So currently we are, we are using. Compliance type must have, which means you must have an object with this cr. So look at this, so this is a gatekeeper constraint, template right, so basically, how we integrate is. You are if you have constraint, template and constraint cr already, so you can use configuration policy to create an unmanaged cluster.

C

So what you just need to do is you just need to embed the constraint template inside the configuration policy and then put it into policy cr which is responsible, distributed across automatic cluster? um That so that's how? How the constraint template is delivered to manage cluster, so you don't you don't have to write any code. You just need to create this policy and embed your constraint template same for the constraint here.

C

Then, the raccoon policy framework will create these two crs on the cluster you on the cluster and then and then gatekeeper is gatekeeper instance. The controllers were taken over and and evaluate these, these two cr. So if you look at details of this cr, this is a sample name. Space constraint, template constraint. I took from uh from gate gatekeeper website, so what it does is it requires any any names. Any this regal basic defines any input.

C

Resource resource needs to have a field called label and inside that label field you need to have a you need to have a have: a have a label uh with name with name with certain name, so the name is being exposed as a configurable parameter called labels and in inside the constraints.

C

So you you're, basically basically applying this constraint, template um I'm here, I'm applying on any namespace and also I restrict the namespace to be these two, these um these, these two namespace, because I I didn't want to apply all um uh want to apply to ordinary space.

C

But if you remove these it will will it will be all namespace and and then I want the actual labels to be gatekeeper, so so for so, for example, if if, if so, if we want to test this at this policy that this gatekeeper constraint, basically, we can just go ahead and create a namespace without label gatekeeper. Then it will be blocked.

C

So let me do do that um here. I believe I've already logged on the login onto the gate, the the manage cluster. So let me double check so here on my manage cluster for this. This deployment I'm actually applying to these two namespace. So if I try to create a namespace.

C

If I do this directly, it will create a namespace without any labels.

C

So you can see here the gatekeeper actually detect that and block the creation of this namespace. So that means the gatekeeper constraint constraint, template that we are that we are apply that we are applied using recom policy is actually in place and and and working.

C

So that's how we uh distribute and apply a gatekeeper constraints, train template from raccoon console by using a recom policy. You don't have to log into each managed cluster and apply it. You just need to do it essentially through right through raccoon control, plane the then it's the the violation part. We want to be able to collect violations.

C

um So if you look at look at the the other two templates that we are embedding in this policy cr, so here as you can see, we create a another configuration policy right. This is for all this scenario. We we name it possibly gatekeeper audit. So what it does is it.

C

It is looking for any cr on the manage cluster called ks required label. This is the constraint that we created in previous object template, but additionally, we are adding additional fields to it. So what it means is we want we want it is. We are using must-have compliance type here. So what it does is we require.

C

We require on your cluster. You need to have a cr called uh ns must have gatekeeper with the this kind and api version. Additionally, we expect the status field with total violation zero.

C

So this is for all this scenario, so the way it works is so once this is, this control constraint is being created on a managed cluster, so gatekeeper audit controller will start to scan your entire cluster and then try to figure out if you have every any pre-existing non-compliance exist. So if there is any, it will actually update the status field. To tell you, oh, there are some some violations pre-exist, um so it will increment the number here so um currently we expect the number to be zero.

C

So if it's not zero, then this this policy will return. Non-Compliant, that's how we collect the result. So if for uh for that's how we collect result for audit and then let's look at the dimension scenario for the dimension scenario same, we are creating another configuration policy and we are looking for event.

C

Events for this constraint, template for this constraint, uh ks, require labels and then was named. Ns must have gold gatekeeper, so basically, so, as you can see here, we are using. Conspiracy. Type must not have so we expect there shouldn't be any events with these annotation unmanaged cluster exist on the manage cluster if it exists. That means there's violations.

C

There's uh there's something wrong, so this policy will return non-compliant. So now, let's come back to the console here. So as I did as previously, I did this right. I tried to create namespace without label. So now, as you can see here, it is for this. This admission template is reported, reporting, non-compliant and it tells oh there's an event exist in the gatekeeper namespace system.

C

um We can click on view, details and take a look at what uh what exactly it is. So, oh there's a this is this is what is violating this part of this policy, so this event exists, but it should it shouldn't, and if you click on your ymo, it is going to pull the the details of this event. So, as you can see, um it tells you admission webhook, uh the the the basically, the the request is being denied by this constraint from gatekeeper.

C

So this is exactly the same error you've seen here in the command line, so this is how we implement uh gatekeeper with recom. So we implement we integrate gatekeeper, not only uh just installing the gatekeeper and then creating the gatekeeper constraint, constraint, template, but we also collect the result for both an audit and animation scenario.

C

um Okay, that's my part for the gatekeeper demo, uh just to quick mention uh thanks to the power of the outer box, configuration policy controller shipped uh with raccoon, so we are able to also, of course, install any operator so, for example, compliance operator here so compliance operator. So if you look at the ammo here this, this policy is also uh you can also find it in positive collection repo. So you can find here so a very similar concept. We use configuration policy to create namespace, create operator group and create subscription.

C

So then the compliance operator is installed on your manage cluster and then you can trigger you can trigger a scan so from using certain profiles. So here, if you look at the ea scan policy here, basically this policy triggers uh trigger this policy triggers an uh essential, a security profile scan on your manage cluster, and we are also using configuration policy to achieve that so by just create using configuration policy to create a scan, set, binding compliance with suite and then collect result back.

C

So this is all done by doing just creating configuration policies, so you don't have to write any code any code. You just need to create a demo of the policy and then you can integrate with compliance operator and as well as gatekeeper.

C

Okay, that's it for me.

D

That's great thanks you, and, and if I could, this is uh kirsten newcomer if I could chime in just for a minute. At the end of our conversation about the client compliance operator, there was in chat a conversation about the difference between the compliance operator and and gatekeeper, and I think both jaya and I weighed in and- and you did a nice job also summarizing that here, so um the compliance operator is specifically focused on on regulatory controls, technical controls um that map to regulatory frameworks.

D

I just want to share just because we're really excited in addition to the e8 profile, which is an essential 8 profile from australia. There are some uh the the compliance profiles available with the openshift compliance operator include some controls that uh meet nist 853 regulatory requirements, including at rel core os, not just the kubernetes layer, and we just shipped uh a profile for that is inspired by the cis kubernetes benchmark so check that out.

D

If, when you get a chance- and I don't know if there are other questions- um I I've kind of been thinking of a few- that I wondered if that, whether they might be helpful to folks um and they kind of tie back to oppa versus uh gatekeeper versus the compliance operator jaya. You mentioned kind of three types of um policies: policy categories- you mentioned security resiliency. You also mentioned software engineering.

D

um I think we have. I I think maybe it would be useful to give some examples of each of them. um Does that make sense.

A

Yeah can I can. I share grab the share again. Please.

A

Do I think you should be able to see my screen right.

D

It's coming there, it is okay,.

A

So uh this is the policy collection, repo and then, if you go down to the bottom of this repo this this page, we have a lot of blogs here, and one of the blogs I wanted to highlight here is this blog.

A

And this blog uh talks about how you can implement policy-based governance, using our configuration management capabilities right and- and here we talk about like kristen- you were asking uh there- are- there- are best practices for three areas that I outlined, which is security, resiliency and software engineering, and this blog goes into the details of how you can of of examples in each of these categories right so, for example, for security, security and regulatory compliance uh of the first one. We list, obviously, is the compliance operator right.

A

um So we talk about uh how that can be configured using rackham, and then we talk about the hcd encryption, so there's a policy that's available to ensure that the kubernetes hct database is encrypted right and and also you can remove the cube admin, because the best practice obviously is to not use a single administrative account right. You want to be able to use different accounts for different users to have accountability and then policies for role-based access, control and sec security, cost constraints, etc right.

A

So these are examples of security, oriented policies, but you can also have policies for resiliency. So, for example, if you want to make sure that the logging operator is installed because you want to be able to collect logs to check the health of clusters etc. So so you can do that and many of the gatekeeper policies um that we have as examples here.

A

I have to go back to the community, so there are gatekeeper policies for ensuring liveness check is in place and make sure you're using the latest container image. You know these kinds of policies, readiness probe is in place.

A

These are all resiliency related policies right, so so those are examples of that and then and then you can also set up policies for uh software engineering, which is things like you know, making making sure that kafka is in place, because that is your standard for streaming right.

A

So so those are the kinds of things you can actually author and, as you was mentioning the key point of this blog. If I go to the conclusion here is right.

A

um The point is, all of these policies can be put in place without programming, and I I wanted to really highlight that, because our built-in config policy controller, as I mentioned earlier, can distribute any cr. So as long as you are able to represent a capability say, for example, as an operator, an operator can be deployed using a cr right and then you can configure the operator as long as it takes configuration as given.

A

It is the resource you can configure it and if it returns the results back as crs, we can process those as well. That's really uh like you are showing in this demo. We were able to integrate with gatekeeper because it met all those characteristics right, so we didn't have to really write code. We just had to author policies, so I think that's the beauty of this approach is it's very easy to put these best practices in place for your data operations. Without writing a lot of code exactly.

D

Yep jaya are there out of the box policies that uh I are available that that in the gatekeeper community, I know you showed us the community for rackham the community branch. What does gatekeeper provide out of the box.

A

um I think many of the policies that we have put here in our community were contributed by our company community of practitioners right, so they have their own community where they based on customer requirements. They are building our policies right and then once they feel that the those policies are stable enough. They then contribute into this community right, okay and then the next progression is.

A

If we find that there are enough customers asking us for a fully supported version of that policy, then we promote that policy to stable right and it becomes part of our dracum product. So that's kind of the threes three-step progression of how we are building our policies and moving them into fully supported versions.

D

That sounds good, um let's see so you answered those examples and um another thing you mentioned early on in the conversation you kind of mentioned. um Well, you used three different words and I just wanted to see. If there was clarity, you said that there are, of course, admission controllers, runtime controllers, and then later you talked about admission and audit, so I wondered, um I think the admission controllers were fairly clear. Could you give us an idea of what some of the runtime controllers are yeah or is that the audit yeah.

A

Yeah, let me go back to my um to sharing my deck.

B

Jaya, we lost your voice again.

A

Can you hear me now yep? Yes, I was just uh going and getting my deck okay, you can see my deck, so I think if you look here, let me go okay, so the policy enforcement points right. um These are the these are the entities that consume the policy and return violations.

A

So this is where you know you have the admission web hook, so anything that is plugged in into that vision. Web hook now. Gatekeeper is one example of that right that is plugged into that now. Somebody could write another web hook. That is processing policies, for example right as long as those um that entity that is configured as a web hook is able to accept its configuration or policy as a kubernetes cr. Then you can use our built-in config policy controller to distribute that to that right.

A

So so that is one example right and then there are communities, obviously is, uh has its own runtime controller. So, for example, when we propagate a policy for a cd encryption that is actually being enforced by kubernetes itself right, we don't have to do an external controller to enforce that right. But then you have certain other controllers, like the certificate management controller is something that we provide and that allows you to detect whether certificates are expiring, whether certificates uh have wild card certificates are being used, etc. Right.

A

Those are policies that you can define and that can be detected using that specific controller and compliance operator is another example right.

D

So, okay, I had thought of things like falco, so um so so it helps me to understand that you mean it more broadly right.

A

Yeah falco is here as well, so that's another example of a runtime controller.

D

Thank you also just really quickly, because I know you and I responded to carolyn in chat, but but maybe we'll just um verbalize for for everybody else as well. uh Carolyn had asked whether openshift is required on all clusters in order to use uh policy-based governance and um no, as as js said in chat, acm itself runs on an open shift cluster, but acm can manage other kubernetes distributions. Besides openshift today, the compliance operator is only running on openshift.

D

It is designed to the the content that's available, um which is written in scaps, secure content, automate security content, automation protocol. This is a a standard, that's required by a number of our customers, which is why the compliance operator is written in scap it. Today uh we only have content for openshift and um the operator mostly works on on openshift.

D

We want to extend it to other coupe clusters in the future um likely or at least share it upstream, to make it easy for others to extend it, um although of course the project is, is open source. So you can do that today, but again, acm itself uh can can manage both openshift and other kubernetes clusters with other kubernetes distros.

C

Okay, just two complements so the the policy framework, the governance policy framework, so the hub piece does require oc, ocp cluster, but the the piece on the manage cluster. We do support other kubernetes, kubernetes um flavor, but also so, but the outer box controllers, for example, the configuration policy controller. It also supports other kubernetes flavors.

C

um So that means, if you want to integrate something that it's not running on on kunis cluster, but uh but it can be integrated using configuration policy. It is you can still do that, and, and also to add, the gatekeeper operator also works on non-kubernetes in work, a non-ocp kubernetes environment. So, of course, if you want to use or olm way to install it, you need to install oem, but you can also install install it without oim. So the upstream version actually supports multiple ways to install gatekeeper.

A

Yeah, those are all excellent points you, I think, um yeah. I wanted to add here like uh if you go to this open cluster management policy collection repo at the very bottom. That's where we've listed all our blogs, so you can take a look at it. Also. This is the open customer management community.

A

So, if you're interested in trying out um this capability, uh that's a good place to start and we have community meetings, um so you can definitely get engaged there and, like you said, you know, for example, if you want to try this out on a non-open shift cluster, um at least on the managed cluster side. Definitely you're welcome to do that.

E

um I have one question: this is shannon.

E

um I have a customer who is running uh currently right now, 311., I notice acm rackham supports um 3.11.200 and above um and uh some of the policies when I install it, um it actually require, like, let's say image, vulnerabilities and it will require container security operator on the managed cluster, and I was just wondering um like what all the policies will install operator for the manage cluster if there isn't, or it is kind of a prereqs- depends on what policies that you're applying to the to the cluster you want.

E

That makes sense like because there there are time that I would not know what my manage cluster has um as an admin. I will want policy a b and c, and then, when I create those policy- and I figure out- oh those clusters doesn't have it uh would would like for gatekeeper. You will deploy the gate keeper for the cluster, but gatekeeper is actually only support, 1.14 and above and so therefore isn't. Does that mean that I cannot have gay keeper policies on the 3.11 cluster.

C

So yeah I can take that. uh So yes, so, as I mentioned previously, there's two pieces, the framework itself and also the policy consumer. So the framework does work on ocp uh on ocp311.

E

C

The policy cons, controller or consumer on match class. It might not support ocp311, so it's really case by case. So so, if you try to create a gatekeeper on 311, then since it's using configuration policy, it will try to apply the those crs. But since those cs doesn't exist even exist on 311, it will return non-combined.

E

Okay, so we'll have to all right, so I have to figure out which one is useful and which one is not um yeah.

A

And I think this is where I think the placement rule applies right. So when you, uh when you are applying a policy to a managed cluster, you do that by using this concept, called placement, and one of the things uh that you can use is you can assign labels to your clusters, and you can say that apply these this policy to a cluster that matches this label right.

A

So so what you could do is you could uh assign labels to your clusters depending upon what version they are, etc right and that will kind of determine you know which policies get applied where.

E

Okay, that's a good idea. Thank you um just thinking out loud because there's real customer questions, uh they um they are on 3 11.. I just want to kind of pick your brain taking the three last three minutes opportunities, um uh and so they they want to follow the needs, um standard, right and so they're running. uh They are moving to four but they're, not yet they're still on three eleven. uh What what are the policies that you would recommend them to start on doing uh with acm if there is any documentation?

E

I can just take the link and read about it.

A

Yeah, um I think the the the blog that I was sharing could be a good place to start. um Let me uh I'll put that link.

E

So nice, I I want to go read I I read them over the weekend. They are super okay, yes, yeah.

D

So jail, we know the com. The compliance operator is not supported on openshift 311., uh there is a hardening guide for openshift 311, but there's not automation uh that I'm aware of does acm have uh checks for openshift 311 like that map to nist 853 or some other nist framework.

D

A

Encryption policy at see the encryption should work on 311 right. Did you support uh openshift4 support? Sorry.

D

Yeah both openshift 311 and you can encrypt cd in both up and shift 311 and open shift. Four. How you do it is slightly different. However, um there are differences in implementation. I just didn't know whether uh any of your other checks yeah.

A

And the other check is the certificate expiration that definitely should work. I think uh that that's another area that you could look at yeah.

D

So shauna, if, if it's for 311 feel this is kirsten, newcomer feel free to shoot me a note. uh Knewcomer redhat.com I'll see what I can do to to help you out. Okay,.

E

Great, thank you so much appreciate it. um Thank you. Thank you.

B

And there's one question from tobias in the chat right now: a question regarding the map item in terms of audit etc. Does this include having a result that could state deny, but not denying the actual resource creation and rather just firing a violation alert to s-I-e-m.

A

The gatekeeper audit scenario, obviously we support right, meaning when I say support that should work. um The roadmap item related to audit was more on the hub being able to generate audit logs for policy lifecycle events and have them routed to sims.

A

So that is something we are we'll work on in the future, but the if you're, talking specifically about the gatekeeper audit scenario, that'll be fine.

B

You should be able.

A

B

Use that this was really really good um and we definitely sounds like we're. Gonna have to have you back um as the ocm community grows and more of these policies are out there. This is just a deep dive here um and very good. You thanks very much for the demo. um It was quite quite good, so thanks thanks again, and um thank you, everybody for taking the time and asking great questions so that I didn't have to um this was this was good and thanks kirsten for for jumping in there take care.

B

Thank you. Everyone, bye-bye.