title: "Active Directory authentication in Tanzu Community Edition" # Title of the blog post.
date: 2022-03-06 # Date of post creation.
# lastmod: 2022-03-03T20:29:39-06:00 # Date when last modified
description: "Deploying and configuring a Tanzu Community Edition Kubernetes cluster to support authentication with Active Directory users and groups by using pinniped and dex" # Description used for search engine.
featured: false # Sets if post is a featured post, making appear on the home page side bar.
draft: false # Sets whether to render this page. Draft of true will not be rendered.
toc: true # Controls if a table of contents should be generated for first-level links automatically.
usePageBundles: true
# menu: main
# featureImage: "file.png" # Sets featured image on blog post.
# featureImageAlt: 'Description of image' # Alternative text for featured image.
# featureImageCap: 'This is the featured image.' # Caption (optional).
Not long ago, I [deployed a Tanzu Community Edition Kubernetes cluster in my homelab](/tanzu-community-edition-k8s-homelab/), and then I fumbled through figuring out how to [log into it from a different device](/logging-in-tce-cluster-from-new-device/) than the one I'd used for deploying the cluster from the `tanzu` cli. That setup works great for playing with Kubernetes in my homelab but I'd love to do some Kubernetes with my team at work and I really need the ability to authenticate multiple users with domain credentials for that.
The TCE team has created a [rather detailed guide](https://tanzucommunityedition.io/docs/latest/vsphere-ldap-config/) for using the [Pinniped](https://pinniped.dev/) and [Dex](https://dexidp.io/docs/kubernetes/) packages[^packages] to provide LDAPS authentication when TCE is connected with the NSX Advanced Load Balancer. This guide got me *most* of the way toward a working solution but I had to make some tweaks along the way (particularly since I'm not using NSX-ALB). I took notes as I worked through it, though, so here I'll share what it actually took to make this work in my environment.
[^packages]: Per VMware, "Pinniped provides the authentication service, which uses Dex to connect to identity providers such as Active Directory."
### Prequisite
In order to put the "Secure" in LDAPS, I need to make sure my Active Directory domain controller is configured for that, and that means also creating a Certificate Authority for issuing certificates. I followed the steps [here](http://vcloud-lab.com/entries/windows-2016-server-r2/configuring-secure-ldaps-on-domain-controller) to get this set up in my homelab. I can then point my browser to `http://win01.lab.bowdre.net/certsrv/certcarc.asp` to download the base64-encoded CA certificate since I'll need that later.
With that sorted, I'm ready to move on to creating a new TCE cluster with an LDAPS identity provider configured.
### Cluster creation
The [cluster deployment steps](/tanzu-community-edition-k8s-homelab/#management-cluster) are very similar to what I did last time so I won't repeat all those instructions here. The only difference is that this time I don't skip past the Identity Management screen; instead, I'll select the LDAPS radio button and get ready to fill out the form.
Before moving on, I can use the **Verify LDAP Configuration** button to quickly confirm that the connection is set up correctly. (I discovered that it doesn't honor the attribute selections made on the previous screen so I have to search for my Common Name (`John Bowdre`) instead of my username (`john`), but this at least lets me verify that the connection and certificate are working correctly.)
I can then click through the rest of the wizard but (as before) I'll stop on the final review page. Despite entering everything correctly in the wizard I'll actually need to make a small edit to the deployment configuration YAML so I make a note of its location and copy it to a file called `tce-mgmt-deploy.yaml` in my working directory so that I can take a look.
#### Editing the cluster spec
Remember that awkward `member:1.2.840.113556.1.4.1941:` attribute from earlier? Here's how it looks within the TCE cluster-defining YAML:
That `:` at the end of the line will cause problems down the road - specifically when the deployment process creates the `dex` app which handles the actual LDAPS authentication piece. Cumulative hours of [troubleshooting](#troubleshooting-notes) (and learning!) eventually revealed to me that something along the way had choked on that trailing colon and inserted this into the `dex` configuration:
That error prevents `dex` from starting correctly so the authentication would never work. I eventually figured out that using the `|` character to define the attribute as a [literal scalar](https://yaml.org/spec/1.2.2/#812-literal-style) would help to get around this issue so I changed the cluster YAML to look like this:
[^member]: Setting this to just `member` would work, but it wouldn't return any nested group memberships. I struggled with this for a long while until I came across a [post from Brian Ragazzi](https://brianragazzi.wordpress.com/2020/05/12/configure-tanzu-kubernetes-grid-to-use-active-directory/) which mentioned using `member:1.2.840.113556.1.4.1941:` instead. This leverages a special [LDAP matching rule OID](https://docs.microsoft.com/en-us/windows/win32/adsi/search-filter-syntax) to expand the nested groups.
#### Deploying the cluster
That's the only thing I need to manually edit so now I can go ahead and create the cluster with:
The [guide I'm following from the TCE site](https://tanzucommunityedition.io/docs/latest/vsphere-ldap-config/) assumes that I'm using NSX-ALB in my environment, but I'm not. So, [as before](/tanzu-community-edition-k8s-homelab/#deploying-kube-vip-as-a-load-balancer), I'll need to deploy [Scott Rosenberg's `kube-vip` Carvel package](https://github.com/vrabbi/tkgm-customizations):
With the load balancer in place, I can follow the TCE instruction to modify the Pinniped and Dex services to switch from the `NodePort` type to the `LoadBalancer` type so they can be easily accessed from outside of the cluster. This process starts by creating a file called `pinniped-supervisor-svc-overlay.yaml` and pasting in the following overlay manifest:
The `-w 0` / `--wrap=0` argument tells `base64` to *not* wrap the encoded lines after a certain number of characters. If you leave this off, the string will get a newline inserted every 76 characters, and those linebreaks would make the string a bit more tricky to work with. Avoid having to clean up the output afterwards by being more specific with the request up front!
{{% /notice %}}
I'll copy the resulting base64 string (which is much longer than the truncated form I'm using here), and paste it into the following command to patch the secret (which will be named after the management cluster name so replace the `tce-mgmt` part as appropriate):
I'll also need to restart the `pinniped-post-deploy-job` job to account for the changes I just made; that's accomplished by simply deleting the existing job. After a few minutes a new job will be spawned automagically. I'll just watch for the new job to be created:
Right now, I've got all the necessary components to support LDAPS authentication with my TCE management cluster but I haven't done anything yet to actually define who should have what level of access. To do that, I'll create a `ClusterRoleBinding`.
I'll toss this into a file I'll call `tanzu-admins-crb.yaml`:
I have a group in Active Directory called `Tanzu-Admins` which contains a group called `vRA-Admins`, and that group contains my user account (`john`). It's a roundabout way of granting access for a single user in this case but it should help to confirm that nested group memberships are being enumerated properly.
After assuming the non-admin context, the next time I try to interact with the cluster it should kick off the LDAPS authentication process. It won't look like anything is happening in the terminal:
[^roles]: You can read up on some other default user-facing roles [here](https://kubernetes.io/docs/reference/access-authn-authz/rbac/#user-facing-roles).
To allow other users to log in this way, I'd need to give them a copy of the non-admin `kubeconfig`, which I can get by running `tanzu management-cluster config get --export-file tce-mgmt-config` to export it into a file named `tce-mgmt-config`. They could use [whatever method they like](https://kubernetes.io/docs/concepts/configuration/organize-cluster-access-kubeconfig/) to merge this in with their existing `kubeconfig`.
{{% notice tip "Tanzu CLI required" %}}
Other users hoping to work with a Tanzu Community Edition cluster will also need to install the Tanzu CLI tool in order to authenticate in this way. Installation instructions can be found [here](https://tanzucommunityedition.io/docs/latest/cli-installation/).
At this point, I've only configured authenticationfor the management cluster - not the workload cluster. The TCE community docs cover what's needed to make this configuration available in the workload cluster as well [here](https://tanzucommunityedition.io/docs/latest/vsphere-ldap-config/#configuration-steps-on-the-workload-cluster). [As before](/tanzu-community-edition-k8s-homelab/#workload-cluster), I created the deployment YAML for the workload cluster by copying the management cluster's deployment YAML and changing the `CLUSTER_NAME` and `VSPHERE_CONTROL_PLANE_ENDPOINT` values accordingly. This time I also deleted all of the `LDAP_*` and `OIDC_*` lines, but made sure to preserve the `IDENTITY_MANAGEMENT_TYPE: ldap` one.
{{% notice note "Create DHCP reservations for control plane nodes" %}}
VMware [points out](https://tanzucommunityedition.io/docs/latest/verify-deployment/#configure-dhcp-reservations-for-the-control-plane-nodes-vsphere-only) that it's important to create DHCP reservations for the IP addresses which were dynamically assigned to the control plane nodes in both the management and workload clusters so be sure to take care of that before getting too involved in "Work".
{{% /notice %}}
[^crb]: Or a different one. In reality, you probably don't want the same users having the same levels of access on both the management and workload clusters. But I stuck with just the one here for now.
### Troubleshooting notes
It took me quite a bit of trial and error to get this far and (being a k8s novice) even more time figuring out how to even troubleshoot the problems I was encountering. So here are a few tips that helped me out.
#### Checking and modifying `dex` configuration
I had a lot of trouble figuring out how to correctly format the `member:1.2.840.113556.1.4.1941:` attribute in the LDAPS config so that it wouldn't get split into multiple attributes due to the trailing colon - and it took me forever to discover that was even the issue. What eventually did the trick for me was learning that I could look at (and modify!) the configuration for the `dex` app with:
This let me make changes on the fly until I got a working configuration and then work backwards from there to format the initial input correctly.
#### Reviewing `dex` logs
Authentication attempts (at least on the LDAPS side of things) will show up in the logs for the `dex` pod running in the `tanzu-system-auth` namespace. This is a great place to look to see if the user isn't being found, credentials are invalid, or the groups aren't being enumerated correctly:
{"level":"info","msg":"performing ldap search OU=LAB,DC=lab,DC=bowdre,DC=net sub (\u0026(objectClass=person)(sAMAccountName=johnny))","time":"2022-03-06T22:29:57Z"}
{"level":"error","msg":"ldap: no results returned for filter: \"(\u0026(objectClass=person)(sAMAccountName=johnny))\"","time":"2022-03-06T22:29:57Z"}
#invalid password
{"level":"info","msg":"performing ldap search OU=LAB,DC=lab,DC=bowdre,DC=net sub (\u0026(objectClass=person)(sAMAccountName=john))","time":"2022-03-06T22:30:45Z"}
{"level":"info","msg":"username \"john\" mapped to entry CN=John Bowdre,OU=Users,OU=BOW,OU=LAB,DC=lab,DC=bowdre,DC=net","time":"2022-03-06T22:30:45Z"}
{"level":"error","msg":"ldap: invalid password for user \"CN=John Bowdre,OU=Users,OU=BOW,OU=LAB,DC=lab,DC=bowdre,DC=net\"","time":"2022-03-06T22:30:45Z"}
# successful login
{"level":"info","msg":"performing ldap search OU=LAB,DC=lab,DC=bowdre,DC=net sub (\u0026(objectClass=person)(sAMAccountName=john))","time":"2022-03-06T22:31:21Z"}
{"level":"info","msg":"username \"john\" mapped to entry CN=John Bowdre,OU=Users,OU=BOW,OU=LAB,DC=lab,DC=bowdre,DC=net","time":"2022-03-06T22:31:21Z"}
{"level":"info","msg":"performing ldap search OU=LAB,DC=lab,DC=bowdre,DC=net sub (\u0026(objectClass=group)(member:1.2.840.113556.1.4.1941:=CN=John Bowdre,OU=Users,OU=BOW,OU=LAB,DC=lab,DC=bowdre,DC=net))","time":"2022-03-06T22:31:21Z"}
I couldn't figure out an elegant way to log out so that I could try authenticating as a different user, but I did discover that information about authenticated sessions get stored in `~/.config/tanzu/pinniped/sessions.yaml`. The sessions expired after a while but until that happens I'm able to keep on interacting with `kubectl` - and not given an option to re-authenticate even if I wanted to.
So this is a pretty basic walkthrough of how I set up my Tanzu Community Edition Kubernetes clusters for Active Directory authentication in my homelab. I feel like I've learned a lot more about TCE specifically and Kubernetes in general through this process, and I'm sure I'll learn more in the future as I keep experimenting with the setup.
