Merge pull request #31 from smallstep/autocert

merge autocert!
pull/37/head
Mariano Cano 5 years ago committed by GitHub
commit 0eb0c3a21b
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23

@ -0,0 +1,28 @@
---
name: Autocert Bug
about: Report a bug you found in autocert
labels: area/autocert bug
---
### Subject of the issue
Describe your issue here
### Environment
* Kubernetes version:
* Cloud provider or hardware configuration:
* OS (e.g., from /etc/os-release):
* Kernel (e.g., `uname -a`):
* Install tools:
* Other:
### Steps to reproduce
Tell us how to reproduce this issue
### Expected behaviour
Tell us what should happen
### Actual behaviour
Tell us what happens instead
### Additional context
Add any other context about the problem here

@ -0,0 +1,11 @@
---
name: Autocert Enhancement
about: Suggest an enhancement to autocert
labels: area/autocert enhancement
---
### What would you like to be added
### Why this is needed

@ -0,0 +1,176 @@
# Installing `autocert`
### Prerequisites
To get started you'll need [`kubectl`](https://kubernetes.io/docs/tasks/tools/install-kubectl/#install-kubectl) and a cluster running kubernetes `1.9` or later with [admission webhooks](https://kubernetes.io/docs/reference/access-authn-authz/extensible-admission-controllers/#admission-webhooks) enabled:
```bash
$ kubectl version --short
Client Version: v1.13.1
Server Version: v1.10.11
$ kubectl api-versions | grep "admissionregistration.k8s.io/v1beta1"
admissionregistration.k8s.io/v1beta1
```
### Install
The easiest way to install `autocert` is to run:
```bash
kubectl run autocert-init -it --rm --image smallstep/autocert-init --restart Never
```
💥 installation complete.
> You might want to [check out what this command does](init/autocert.sh) before running it.
## Manual install
To install manually you'll need to [install step](https://github.com/smallstep/cli#installing) version `0.8.3` or later.
```
$ step version
Smallstep CLI/0.8.3 (darwin/amd64)
Release Date: 2019-01-16 01:46 UTC
```
### Create a CA
Set your `STEPPATH` to a working directory where we can stage our CA artifacts before we push them to kubernetes. You can delete this directory once installation is complete.
```
$ export STEPPATH=$(mktemp -d /tmp/step.XXX)
$ step path
/tmp/step.0kE
```
Run `step ca init` to generate a root certificate and CA configuration for your cluster. You'll be prompted for a password that will be used to encrypt key material.
```
$ step ca init \
--name Autocert \
--dns "ca.step.svc.cluster.local,127.0.0.1" \
--address ":4443" \
--provisioner admin \
--with-ca-url "ca.step.svc.cluster.local"
```
For older versions of `step` run this command without the flags.
Add provisioning credentials for use by `autocert`. You'll be prompted for a password for `autocert`.
```
$ step ca provisioner add autocert --create
```
For older versions of `step`:
* Run `step ca init` and follow prompts
* Edit `$(step path)/config/ca.json` and change base paths to `/home/step`
* Edit `$(step path)/config/defaults.json` to change base paths to `/home/step` and remove port from CA URL
```
$ sed -i "" "s|$(step path)|/home/step/.step|g" $(step path)/config/ca.json
$ sed -i "" "s|$(step path)|/home/step/.step|g" $(step path)/config/defaults.json
$ sed -i "" "s|ca.step.svc.cluster.local:4443|ca.step.svc.cluster.local|" $(step path)/config/defaults.json
```
### Install the CA in Kubernetes
We'll be creating a new kubernetes namespace and setting up some RBAC rules during installation. You'll need appropriate permissions in your cluster (e.g., you may need to be cluster-admin). GKE, in particular, does not give the cluster owner these rights by default. You can give yourself cluster-admin rights on GKE by running:
```bash
kubectl create clusterrolebinding cluster-admin-binding \
--clusterrole cluster-admin \
--user $(gcloud config get-value account)
```
We'll install our CA and the `autocert` controller in the `step` namespace.
```
$ kubectl create namespace step
```
To install the CA we need to configmap the CA certificates, signing keys, and configuration artifacts. Note that key material is encrypted so we don't need to use secrets.
```
$ kubectl -n step create configmap config --from-file $(step path)/config
$ kubectl -n step create configmap certs --from-file $(step path)/certs
$ kubectl -n step create configmap secrets --from-file $(step path)/secrets
```
But we will need to create secrets for the CA and autocert to decrypt their keys:
```
$ kubectl -n step create secret generic ca-password --from-literal password=<ca-password>
$ kubectl -n step create secret generic autocert-password --from-literal password=<autocert-password>
```
Where `<ca-password>` is the password you entered during `step ca init` and `<autocert-password>` is the password you entered during `step ca provisioner add`.
Next, we'll install the CA.
```
$ kubectl apply -f https://raw.githubusercontent.com/smallstep/certificates/master/autocert/install/01-step-ca.yaml
```
Once you've done this you can delete the temporary `$STEPPATH` directory and `unset STEPPATH` (though you may want to retain it as a backup).
### Install `autocert` in Kubernetes
Install the `autocert` controller.
```
$ kubectl apply -f https://raw.githubusercontent.com/smallstep/certificates/master/autocert/install/02-autocert.yaml
```
Autocert creates secrets containing single-use bootstrap tokens for pods to authenticate with the CA and obtain a certificate. The tokens are automatically cleaned up after they expire. To do this, `autocert` needs permission to create and delete secrets in your cluster.
If you have RBAC enabled in your cluster, apply `rbac.yaml` to give `autocert` these permissions.
```
$ kubectl apply -f https://raw.githubusercontent.com/smallstep/certificates/master/autocert/install/03-rbac.yaml
```
Finally, register the `autocert` mutation webhook with kubernetes.
```
$ cat <<EOF | kubectl apply -f -
apiVersion: admissionregistration.k8s.io/v1beta1
kind: MutatingWebhookConfiguration
metadata:
name: autocert-webhook-config
labels: {app: autocert}
webhooks:
- name: autocert.step.sm
clientConfig:
service:
name: autocert
namespace: step
path: "/mutate"
caBundle: $(cat $(step path)/certs/root_ca.crt | base64)
rules:
- operations: ["CREATE"]
apiGroups: [""]
apiVersions: ["v1"]
resources: ["pods"]
failurePolicy: Ignore
namespaceSelector:
matchLabels:
autocert.step.sm: enabled
EOF
```
### Check your work
If everything worked you should have CA and controller pods running in the `step` namespace and your webhook configuration should be installed:
```
$ kubectl -n step get pods
NAME READY STATUS RESTARTS AGE
ca-7577d7d667-vtfq5 1/1 Running 0 1m
controller-86bd99bd96-s9zlc 1/1 Running 0 28s
$ kubectl get mutatingwebhookconfiguration
NAME CREATED AT
autocert-webhook-config 2019-01-17T22:57:57Z
```

@ -0,0 +1,413 @@
![Autocert architecture diagram](https://raw.githubusercontent.com/smallstep/certificates/autocert/autocert/autocert-logo.png)
# Autocert
[![GitHub release](https://img.shields.io/github/release/smallstep/certificates.svg)](https://github.com/smallstep/certificates/releases)
[![CA Image](https://images.microbadger.com/badges/image/smallstep/step-ca.svg)](https://microbadger.com/images/smallstep/step-ca)
[![Go Report Card](https://goreportcard.com/badge/github.com/smallstep/certificates)](https://goreportcard.com/report/github.com/smallstep/certificates)
[![GitHub stars](https://img.shields.io/github/stars/smallstep/certificates.svg?style=social)](https://github.com/smallstep/certificates/stargazers)
[![Twitter followers](https://img.shields.io/twitter/follow/smallsteplabs.svg?label=Follow&style=social)](https://twitter.com/intent/follow?screen_name=smallsteplabs)
**Autocert** is a kubernetes add-on that automatically injects TLS/HTTPS certificates into your containers.
To get a certificate **simply annotate your pods** with a name. An X.509 (TLS/HTTPS) certificate is automatically created and mounted at `/var/run/autocert.step.sm/` along with a corresponding private key and root certificate (everything you need for [mTLS](#motivation)).
We ❤️ feedback. Please [report bugs](https://github.com/smallstep/certificates/issues/new?template=autocert_bug.md) & [suggest enhancements](https://github.com/smallstep/certificates/issues/new?template=autocert_enhancement.md). [Fork](https://github.com/smallstep/certificates/fork) and send a PR. [Give us a ⭐](https://github.com/smallstep/certificates/stargazers) if you like what we're doing.
![Autocert demo gif](https://raw.githubusercontent.com/smallstep/certificates/autocert/autocert/demo.gif)
## Motivation
`Autocert` exists to **make it easy to use mTLS** ([mutual TLS](examples/hello-mtls/README.md#mutual-tls)) to **improve security** within a cluster and to **secure communication into, out of, and between kubernetes clusters**.
TLS (and HTTPS, which is HTTP over TLS) provides _authenticated encryption_: an _identity dialtone_ and _end-to-end encryption_ for your workloads. It's like a secure line with caller ID. This has all sorts of benefits: better security, compliance, and easier auditability for starters. It **makes workloads identity-aware**, improving observability and enabling granular access control. Perhaps most compelling, mTLS lets you securely communicate with workloads running anywhere, not just inside kubernetes.
![Connect with mTLS diagram](https://raw.githubusercontent.com/smallstep/certificates/autocert/autocert/connect-with-mtls.png)
Unlike VPNs & SDNs, deploying and scaling mTLS is pretty easy. You're (hopefully) already using TLS, and your existing tools and standard libraries will provide most of what you need. If you know how to operate DNS and reverse proxies, you know how to operate mTLS infrastructure.
There's just one problem: **you need certificates issued by your own certificate authority (CA)**. Building and operating a CA, issuing certificates, and making sure they're renewed before they expire is tricky. `Autocert` does all of this for you.
## Features
First and foremost, `autocert` is easy. You can **get started in minutes**.
`Autocert` uses [`step certificates`](https://github.com/smallstep/certificates) to generate keys and issue certificates. This process is secure and automatic, all you have to do is [install autocert](#install) and [annotate your pods](#enable-autocert-per-namespace).
Features include:
* A fully featured private **certificate authority** (CA) for workloads running on kubernetes and elsewhere
* [RFC5280](https://tools.ietf.org/html/rfc5280) and [CA/Browser Forum](https://cabforum.org/baseline-requirements-documents/) compliant certificates that work **for TLS**
* Namespaced installation into the `step` namespace so it's **easy to lock down** your CA
* Short-lived certificates with **fully automated** enrollment and renewal
* Private keys are never transmitted across the network and aren't stored in `etcd`
Because `autocert` is built on [`step certificates`](https://github.com/smallstep/certificates) you can easily [extend access](#connecting-from-outside-the-cluster) to developers, endpoints, and workloads running outside your cluster, too.
## Getting Started
> ⚠️ Warning: *this project is in **ALPHA**. DON'T use it for anything mission critical. EXPECT breaking changes in minor revisions with little or no warning. PLEASE [provide feedback](https://github.com/smallstep/certificates/issues/new?template=autocert_enhancement.md).*
### Prerequisites
All you need to get started is [`kubectl`](https://kubernetes.io/docs/tasks/tools/install-kubectl/#install-kubectl) and a cluster running kubernetes `1.9` or later with [admission webhooks](https://kubernetes.io/docs/reference/access-authn-authz/extensible-admission-controllers/#admission-webhooks) enabled:
```bash
$ kubectl version --short
Client Version: v1.13.1
Server Version: v1.10.11
$ kubectl api-versions | grep "admissionregistration.k8s.io/v1beta1"
admissionregistration.k8s.io/v1beta1
```
### Install
To install `autocert` run:
```bash
kubectl run autocert-init -it --rm --image smallstep/autocert-init --restart Never
```
💥 installation complete.
> You might want to [check out what this command does](init/autocert.sh) before running it. You can also [install `autocert` manually](INSTALL.md#manual-install) if that's your style.
## Usage
Using `autocert` is also easy:
* Enable `autocert` for a namespace by labelling it with `autocert.step.sm=enabled`, then
* Inject certificates into containers by annotating pods with `autocert.step.sm/name: <name>`
### Enable autocert (per namespace)
To enable `autocert` for a namespace it must be labelled `autocert.step.sm=enabled`.
To label the `default` namespace run:
```bash
kubectl label namespace default autocert.step.sm=enabled
```
To check which namespaces have `autocert` enabled run:
```bash
$ kubectl get namespace -L autocert.step.sm
NAME STATUS AGE AUTOCERT.STEP.SM
default Active 59m enabled
...
```
### Annotate pods to get certificates
To get a certificate you need to tell `autocert` your workload's name using the `autocert.step.sm/name` annotation (this name will appear as the X.509 common name and SAN).
Let's deploy a [simple mTLS server](examples/hello-mtls/go/server.go) named `hello-mtls.default.svc.cluster.local`:
```yaml
cat <<EOF | kubectl apply -f -
apiVersion: apps/v1
kind: Deployment
metadata: {name: hello-mtls, labels: {app: hello-mtls}}
spec:
replicas: 1
selector: {matchLabels: {app: hello-mtls}}
template:
metadata:
annotations:
# AUTOCERT ANNOTATION HERE -v ###############################
autocert.step.sm/name: hello-mtls.default.svc.cluster.local #
# AUTOCERT ANNOTATION HERE -^ ###############################
labels: {app: hello-mtls}
spec:
containers:
- name: hello-mtls
image: smallstep/hello-mtls-server-go:latest
EOF
```
In our new container we should find a certificate, private key, and root certificate mounted at `/var/run/autocert.step.sm`:
```bash
$ export HELLO_MTLS=$(kubectl get pods -l app=hello-mtls -o jsonpath={$.items[0].metadata.name})
$ kubectl exec -it $HELLO_MTLS -c hello-mtls -- ls /var/run/autocert.step.sm
root.crt site.crt site.key
```
We're done. Our container has a certificate, issued by our CA, which `autocert` will automatically renew.
✅ Certificates.
## Hello mTLS
It's easy to deploy certificates using `autocert`, but it's up to you to use them correctly. To get you started, [`hello-mtls`](examples/hello-mtls) demonstrates the right way to use mTLS with various tools and languages (contributions welcome :). If you're a bit fuzzy on how mTLS works, [the `hello-mtls` README](examples/hello-mtls/README.md) is a great place to start.
To finish out this tutorial let's keep things simple and try `curl`ing the server we just deployed from inside and outside the cluster.
### Connecting from inside the cluster
First, let's expose our workload to the rest of the cluster using a service:
```
kubectl expose deployment hello-mtls --port 443
```
Now let's deploy a client, with its own certificate, that [`curl`s our server in a loop](examples/hello-mtls/curl/client.sh):
```yaml
cat <<EOF | kubectl apply -f -
apiVersion: apps/v1
kind: Deployment
metadata: {name: hello-mtls-client, labels: {app: hello-mtls-client}}
spec:
replicas: 1
selector: {matchLabels: {app: hello-mtls-client}}
template:
metadata:
annotations:
# AUTOCERT ANNOTATION HERE -v ######################################
autocert.step.sm/name: hello-mtls-client.default.pod.cluster.local #
# AUTOCERT ANNOTATION HERE -^ ######################################
labels: {app: hello-mtls-client}
spec:
containers:
- name: hello-mtls-client
image: smallstep/hello-mtls-client-curl:latest
env: [{name: HELLO_MTLS_URL, value: https://hello-mtls.default.svc.cluster.local}]
EOF
```
> Note that **the authority portion of the URL** (the `HELLO_MTLS_URL` env var) **matches the name of the server we're connecting to** (both are `hello-mtls.default.svc.cluster.local`). That's required for standard HTTPS and can sometimes require some DNS trickery.
Once deployed we should start seeing the client log responses from the server [saying hello](examples/hello-mtls/go/server.go#L71-L72):
```
$ export HELLO_MTLS_CLIENT=$(kubectl get pods -l app=hello-mtls-client -o jsonpath={$.items[0].metadata.name})
$ kubectl logs $HELLO_MTLS_CLIENT -c hello-mtls-client
Thu Feb 7 23:35:23 UTC 2019: Hello, hello-mtls-client.default.pod.cluster.local!
Thu Feb 7 23:35:28 UTC 2019: Hello, hello-mtls-client.default.pod.cluster.local!
```
For kicks, let's `exec` into this pod and try `curl`ing ourselves:
```
$ kubectl exec $HELLO_MTLS_CLIENT -c hello-mtls-client -- curl -sS \
--cacert /var/run/autocert.step.sm/root.crt \
--cert /var/run/autocert.step.sm/site.crt \
--key /var/run/autocert.step.sm/site.key \
https://hello-mtls.default.svc.cluster.local
Hello, hello-mtls-client.default.pod.cluster.local!
```
✅ mTLS inside cluster.
### Connecting from outside the cluster
Connecting from outside the cluster is a bit more complicated. We need to handle DNS and obtain a certificate ourselves. These tasks were handled automatically inside the cluster by kubernetes and `autocert`, respectively.
That said, because our server uses mTLS **only clients that have a certificate issued by our certificate authority will be allowed to connect**. That means it can be safely and easily exposed directly to the public internet using a [LoadBalancer service type](https://kubernetes.io/docs/concepts/services-networking/service/#loadbalancer):
```
kubectl expose deployment hello-mtls --name=hello-mtls-lb --port=443 --type=LoadBalancer
```
To connect we need a certificate. There are a [couple](RUNBOOK.md#federation) [different](RUNBOOK.md#multiple-intermediates) [ways](RUNBOOK.md#exposing-the-ca) to get one, but for simplicity we'll just forward a port.
```
kubectl -n step port-forward $(kubectl -n step get pods -l app=ca -o jsonpath={$.items[0].metadata.name}) 4443:4443
```
In another window we'll use `step` to grab the root certificate, generate a key pair, and get a certificate.
> To follow along you'll need to [`install step`](https://github.com/smallstep/cli#installing) if you haven't already. You'll also need your admin password and CA fingerprint, which were output during installation (see [here](RUNBOOK.md#recover-admin-and-ca-password) and [here](RUNBOOK.md#recompute-root-certificate-fingerprint) if you already lost them :).
```bash
$ export CA_POD=$(kubectl -n step get pods -l app=ca -o jsonpath={$.items[0].metadata.name})
$ step ca root root.crt --ca-url https://127.0.0.1:4443 --fingerprint <fingerprint>
$ step ca certificate mike mike.crt mike.key --ca-url https://127.0.0.1:4443 --root root.crt
✔ Key ID: H4vH5VfvaMro0yrk-UIkkeCoPFqEfjF6vg0GHFdhVyM (admin)
✔ Please enter the password to decrypt the provisioner key: 0QOC9xcq56R1aEyLHPzBqN18Z3WfGZ01
✔ CA: https://127.0.0.1:4443/1.0/sign
✔ Certificate: mike.crt
✔ Private Key: mike.key
```
Now we can simply `curl` the service:
> If you're using minikube or docker for mac the load balancer's "IP" might be `localhost`, which won't work. In that case, simply `export HELLO_MTLS_IP=127.0.0.1` and try again.
```
$ export HELLO_MTLS_IP=$(kubectl get svc hello-mtls-lb -ojsonpath={$.status.loadBalancer.ingress[0].ip})
$ curl --resolve hello-mtls.default.svc.cluster.local:443:$HELLO_MTLS_IP \
--cacert root.crt \
--cert mike.crt \
--key mike.key \
https://hello-mtls.default.svc.cluster.local
Hello, mike!
```
> Note that we're using `--resolve` to tell `curl` to override DNS and resolve the name in our workload's certificate to its public IP address. In a real production infrastructure you could configure DNS manually, or you could propagate DNS to workloads outside kubernetes using something like [ExternalDNS](https://github.com/kubernetes-incubator/external-dns).
✅ mTLS outside cluster.
<!--- TODO: CTA or Further Reading... Move "How it works" maybe? Or put this below that? --->
### Cleanup & uninstall
To clean up after running through the tutorial remove the `hello-mtls` and `hello-mtls-client` deployments and services:
```
kubectl delete deployment hello-mtls
kubectl delete deployment hello-mtls-client
kubectl delete service hello-mtls
kubectl delete service hello-mtls-lb
```
See the runbook for instructions on [uninstalling `autocert`](RUNBOOK.md#uninstalling).
## How it works
### Architecture
`Autocert` is an [admission webhook](https://kubernetes.io/docs/reference/access-authn-authz/extensible-admission-controllers/#admission-webhooks) that intercepts and patches pod creation requests with [some YAML](install/02-autocert.yaml#L26-L44) to inject an [init container](bootstrapper/) and [sidecar](renewer/) that handle obtaining and renewing certificates, respectively.
![Autocert architecture diagram](https://raw.githubusercontent.com/smallstep/certificates/autocert/autocert/autocert-arch.png)
### Enrollment & renewal
It integrates with [`step certificates`](https://github.com/smallstep/certificates) and uses the [one-time token bootstrap protocol](https://smallstep.com/blog/step-certificates.html#automated-certificate-management) from that project to mutually authenticate a new pod with your certificate authority, and obtain a certificate.
![Autocert bootstrap protocol diagram](https://raw.githubusercontent.com/smallstep/certificates/autocert/autocert/autocert-bootstrap.png)
Tokens are [generated by the admission webhook](controller/provisioner.go#L46-L72) and [transmitted to the injected init container via a kubernetes secret](controller/main.go#L91-L125). The init container [uses the one-time token](bootstrapper/bootstrapper.sh) to obtain a certificate. A sidecar is also installed to [renew certificates](renewer/Dockerfile#L8) before they expire. Renewal simply uses mTLS with the CA.
## Further Reading
* We tweet [@smallsteplabs](https://twitter.com/smallsteplabs)
* Read [our blog](https://smallstep.com/blog)
* Check out the [runbook](RUNBOOK.md)
* Check out [`step` CLI](https://github.com/smallstep/cli)
## Questions
#### Wait, so any pod can get a certificate with any identity? How is that secure?
1. Don't give people `kubectl` access to your production clusters
2. Use a deploy pipeline based on `git` artifacts
3. Enforce code review on those `git` artifacts
If that doesn't work for you, or if you have a better idea, we'd love to hear! Please [open an issue](https://github.com/smallstep/certificates/issues/new?template=autocert_enhancement.md)!
#### Why do I have to tell you the name to put in a certificate? Why can't you automatically bind service names?
Mostly because monitoring the API server to figure out which services are associated with which workloads is complicated and somewhat magical. And it might not be what you want.
That said, we're not totally opposed to this idea. If anyone has strong feels and a good design please [open an issue](https://github.com/smallstep/certificates/issues/new?template=autocert_enhancement.md).
#### Doesn't kubernetes already ship with a certificate authority?
Yes, it uses [a bunch of CAs](https://jvns.ca/blog/2017/08/05/how-kubernetes-certificates-work/) for different sorts of control plane communication. Technically, kubernetes doesn't _come with_ a CA. It has integration points that allow you to use any CA (e.g., [Kubernetes the hard way](https://github.com/kelseyhightower/kubernetes-the-hard-way) [uses CFSSL](https://github.com/kelseyhightower/kubernetes-the-hard-way/blob/2983b28f13b294c6422a5600bb6f14142f5e7a26/docs/02-certificate-authority.md). You could use [`step certificates`](https://github.com/smallstep/certificates), which `autocert` is based on, instead.
In any case, these CAs are meant for control plane communication. You could use them for your service-to-service data plane, but it's probably not a good idea.
#### What permissions does `autocert` require in my cluster and why?
`Autocert` needs permission to create and delete secrets cluster-wide. You can [check out our RBAC config here](install/03-rbac.yaml). These permissions are needed in order to transmit one-time tokens to workloads using secrets, and to clean up afterwards. We'd love to scope these permissions down further. If anyone has any ideas please [open an issue](https://github.com/smallstep/certificates/issues/new?template=autocert_enhancement.md).
#### Why does `autocert` create secrets?
The `autocert` admission webhook needs to securely transmit one-time bootstrap tokens to containers. This could be accomplished without using secrets. The webhook returns a [JSONPatch](https://tools.ietf.org/html/rfc6902) response that's applied to the pod spec. This response could patch the literal token value into our init container's environment.
Unfortunately, the kubernetes API server does not authenticate itself to admission webhooks by default, and configuring it to do so [requires passing a custom config file](https://kubernetes.io/docs/reference/access-authn-authz/extensible-admission-controllers/#authenticate-apiservers) at apiserver startup. This isn't an option for everyone (e.g., on GKE) so we opted not to rely on it.
Since our webhook can't authenticate callers, including bootstrap tokens in patch responses would be dangerous. By using secrets an attacker can still trick `autocert` into generating superflous bootstrap tokens, but they'd also need read access to cluster secrets to do anything with them.
Hopefully this story will improve with time.
#### Why not use kubernetes service accounts instead of bootstrap tokens?
Great idea! This should be pretty easy to add. However, existing service accounts are [somewhat broken](https://github.com/kubernetes/community/pull/1460) for this use case. The upcoming [TokenRequest API](https://github.com/kubernetes/kubernetes/issues/58790) should fix most of these issues.
TODO: Link to issue for people who want this.
#### Too. many. containers. Why do you need to install an init container and sidecar?
We don't. It's just easier for you. Your containers can generate key pairs, exchange them for certificates, and manage renewals themselves. This is pretty easy if you [install `step`](https://github.com/smallstep/cli#installing) in your containers, or integrate with our [golang SDK](https://godoc.org/github.com/smallstep/certificates/ca). To support this we'd need to add the option to inject a bootstrap token without injecting these containers.
TODO: Link to issue for people who want this.
That said, the init container and sidecar are both super lightweight.
#### Why are keys and certificates managed via volume mounts? Why not use a secret or some custom resource?
Because, by default, kubernetes secrets are stored in plaintext in `etcd` and might even be transmitted unencrypted across the network. Even if secrets were properly encrypted, transmitting a private key across the network violates PKI best practices. Key pairs should always be generated where they're used, and private keys should never be known by anyone but their owners.
That said, there are use cases where a certificate mounted in a secret resource is desirable (e.g., for use with a kubernetes `Ingress`). We may add support for this in the future. However, we think the current method is easier and a better default.
TODO: Link to issue for people who want this.
#### Why not use kubernetes CSR resources for this?
It's harder and less secure. If any good and simple design exists for securely automating CSR approval using this resource we'd love to see it!
#### How is this different than [`cert-manager`](https://github.com/jetstack/cert-manager)
`Cert-manager` is a great project. But it's design is focused on managing Web PKI certificates issued by [Let's Encrypt's](https://letsencrypt.org/) public certificate authority. These certificates are useful for TLS ingress from web browsers. `Autocert` is different. It's purpose-built to manage certificates issued by your own private CA to support the use of mTLS for internal communication (e.g., service-to-service).
#### What sorts of keys are issued and how often are certificates rotated?
`Autocert` builds on `step certificates` which issues ECDSA certificates using the P256 curve with ECDSA-SHA256 signatures by default. If this is all Greek to you, rest assured these are safe, sane, and modern defaults that are suitable for the vast majority of environments.
#### What crypto library is under the hood?
https://golang.org/pkg/crypto/
## Building
This project is based on four docker containers. They use [multi-stage builds](https://docs.docker.com/develop/develop-images/multistage-build/) so all you need in order to build them is `docker`.
> Caveat: the `controller` container uses [`dep`](https://github.com/golang/dep) and `dep init` isn't run during the build. You'll need to run `dep init` in the `controller/` subdirectory prior to building, and you'll need to run `dep ensure -update` if you change any dependencies.
Building `autocert-controller` (the admission webhook):
```
cd controller
docker build -t smallstep/autocert-controller:latest .
```
Building `autocert-bootstrapper` (the init container that generates a key pair and exchanges a bootstrap token for a certificate):
```
cd bootstrapper
docker build -t smallstep/autocert-bootstrapper:latest .
```
Building `autocert-renewer` (the sidecar that renews certificates):
```
cd renewer
docker build -t smallstep/autocert-renewer:latest .
```
Building `autocert-init` (the install script):
```
cd init
docker build -t smallstep/autocert-init:latest .
```
If you build your own containers you'll probably need to [install manually](INSTALL.md). You'll also need to adjust which images are deployed in the [deployment yaml](install/02-autocert.yaml).
## Contributing
If you have improvements to `autocert`, send us your pull requests! For those just getting started, GitHub has a [howto](https://help.github.com/articles/about-pull-requests/). A team member will review your pull requests, provide feedback, and merge your changes. In order to accept contributions we do need you to [sign our contributor license agreement](https://cla-assistant.io/smallstep/certificates).
If you want to contribute but you're not sure where to start, take a look at the [issues with the "good first issue" label](https://github.com/smallstep/certificates/issues?q=is%3Aopen+label%3A%22good+first+issue%22+label%3Aarea%2Fautocert). These are issues that we believe are particularly well suited for outside contributions, often because we probably won't get to them right now. If you decide to start on an issue, leave a comment so that other people know that you're working on it. If you want to help out, but not alone, use the issue comment thread to coordinate.
If you've identified a bug or have ideas for improving `autocert` that you don't have time to implement, we'd love to hear about them. Please open an issue to [report a bug](https://github.com/smallstep/certificates/issues/new?template=autocert_bug.md) or [suggest an enhancement](https://github.com/smallstep/certificates/issues/new?template=autocert_enhancement.md)!
## License
Copyright 2019 Smallstep Labs
Licensed under [the Apache License, Version 2.0](https://github.com/smallstep/certificates/blob/master/LICENSE)

@ -0,0 +1,122 @@
# Runbook
## Common admin tasks
#### Recover `admin` and CA password
```
kubectl -n step get secret ca-password -o jsonpath='{$.data.password}' | base64 -D
```
#### Recover `autocert` password
```
kubectl -n step get secret autocert-password -o jsonpath='{$.data.password}' | base64 -D
```
#### Recompute root certificate fingerprint
```
export CA_POD=$(kubectl -n step get pods -l app=ca -o jsonpath={$.items[0].metadata.name})
kubectl -n step exec -it $CA_POD step certificate fingerprint /home/step/.step/certs/root_ca.crt
```
> Tip: Some slight fanciness is necessary to trim this string if you want to put it into an environment variable:
>
> ```
> export FINGERPRINT="$(kubectl -n step exec -it $CA_POD step certificate fingerprint /home/step/.step/certs/root_ca.crt | tr -d '[:space:]')"
> ```
#### Inspect a certificate
```
kubectl exec -it <pod> -c autocert-renewer -- step certificate inspect /var/run/autocert.step.sm/site.crt
```
#### Labelling a namespace (enabling `autocert` for a namespace)
To enable `autocert` for a namespace it must be labelled. To label an existing namespace run:
```
kubectl label namespace <namespace> autocert.step.sm=enabled
```
#### Checking which namespaces are labelled
```
kubectl get namespace -L autocert.step.sm
```
#### Removing a label from a namespace (disabling `autocert` for a namespace)
```
kubectl label namespace <namespace> autocert.step.sm-
```
#### Naming considerations
Use hostnames. Must be global. Everyone who connects to the service using mTLS must use the same hostname. For internal communication it's easy enough to use the FQDN of a service. For stuff you expose publicly you'll need to manage DNS yourself...
In any case, the critical invariant is: ...
Diagram here?
#### Cleaning up one-time token secrets
```
for ns in $(kubectl get namespace --selector autocert.step.sm=enabled -o jsonpath='{$.items[*].metadata.name}'); do
kubectl -n "$ns" delete secrets --selector="autocert.step.sm/token=true"
done
```
### TODO:
* Change admin password
* Change autocert password
* Federating with another CA
* DNS tips and tricks
* Multiple SANs
* Getting rid of the sidecar
* Getting logs from the CA (certificates weren't issued)
* Getting logs from the init container / renewer (didn't start properly)
* Adjusting certificate expiration (default 24h)
* Remove label
* Clean up secrets
* Naming considerations (maybe this should be in hello-mtls)
## Federation
TODO: Example of federating a CA running in kubernetes with another CA.
For now, see https://smallstep.com/blog/step-v0.8.3-federation-root-rotation.html
## Multiple intermediates
TODO: Example of creating an additional intermediate signing certificate off of our kubernetes root CA.
For now, see https://smallstep.com/docs/cli/ca/init/ (specifically, the `--root` flag)
## Exposing the CA
Beware that the CA exposes an unauthenticated endpoint that lists your configured provisioners and their encrypted private keys. For this reason, you may not want to expose it directly to the public internet.
## Uninstalling
To uninstall `autocert` completely simply delete the mutating webhook configuration, the `step` namespace and the `autocert` RBAC artifacts:
```
kubectl delete mutatingwebhookconfiguration autocert-webhook-config
kubectl delete namespace step
kubectl delete clusterrolebinding autocert-controller
kubectl delete clusterrole autocert-controller
```
Remove any namespace labels and clean up any stray secrets that `autocert` hasn't cleaned up yet:
```
for ns in $(kubectl get namespace --selector autocert.step.sm=enabled -o jsonpath='{$.items[*].metadata.name}'); do
kubectl label namespace "$ns" autocert.step.sm-
kubectl -n "$ns" delete secrets --selector="autocert.step.sm/token=true"
done
```
Any remaining sidecar containers will go away once you remove annotations and re-deploy your workloads.

Binary file not shown.

After

Width:  |  Height:  |  Size: 566 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 396 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 112 KiB

@ -0,0 +1,10 @@
FROM smallstep/step-cli:0.8.3
USER root
ENV CRT="/var/run/autocert.step.sm/site.crt"
ENV KEY="/var/run/autocert.step.sm/site.key"
ENV STEP_ROOT="/var/run/autocert.step.sm/root.crt"
COPY bootstrapper.sh /home/step/
RUN chmod +x /home/step/bootstrapper.sh
CMD ["/home/step/bootstrapper.sh"]

@ -0,0 +1,7 @@
#!/bin/sh
# Download the root certificate and set permissions
step ca certificate $COMMON_NAME $CRT $KEY
chmod 644 $CRT $KEY
step ca root $STEP_ROOT

Binary file not shown.

After

Width:  |  Height:  |  Size: 669 KiB

@ -0,0 +1,19 @@
# build stage
FROM golang:alpine AS build-env
RUN apk update && apk upgrade && \
apk add --no-cache git
RUN go get -u github.com/golang/dep/cmd/dep
WORKDIR $GOPATH/src/github.com/step-certificates-k8s/controller
# copy dep files and run dep separately from code for better caching
COPY Gopkg.toml Gopkg.lock ./
RUN dep ensure --vendor-only
COPY . ./
RUN go build -o /server .
# final stage
FROM smallstep/step-cli:0.8.3
ENV STEPPATH="/home/step/.step"
ENV PWDPATH="/home/step/password/password"
ENV CONFIGPATH="/home/step/autocert/config.yaml"
COPY --from=build-env /server .
ENTRYPOINT ./server $CONFIGPATH

@ -0,0 +1,124 @@
package main
import (
"crypto/tls"
"crypto/x509"
"fmt"
"io/ioutil"
"net"
"net/http"
"os"
"strings"
"time"
)
const (
serviceAccountToken = "/var/run/secrets/kubernetes.io/serviceaccount/token"
serviceAccountCACert = "/var/run/secrets/kubernetes.io/serviceaccount/ca.crt"
)
// Client is minimal kubernetes client interface
type Client interface {
Do(req *http.Request) (*http.Response, error)
GetRequest(url string) (*http.Request, error)
PostRequest(url, body, contentType string) (*http.Request, error)
DeleteRequest(url string) (*http.Request, error)
Host() string
}
type k8sClient struct {
host string
token string
httpClient *http.Client
}
func (kc *k8sClient) GetRequest(url string) (*http.Request, error) {
if !strings.HasPrefix(url, kc.host) {
url = fmt.Sprintf("%s/%s", kc.host, url)
}
req, err := http.NewRequest("GET", url, nil)
if err != nil {
return nil, err
}
if len(kc.token) > 0 {
req.Header.Set("Authorization", "Bearer "+kc.token)
}
return req, nil
}
func (kc *k8sClient) PostRequest(url string, body string, contentType string) (*http.Request, error) {
if !strings.HasPrefix(url, kc.host) {
url = fmt.Sprintf("%s/%s", kc.host, url)
}
req, err := http.NewRequest("POST", url, strings.NewReader(body))
if err != nil {
return nil, err
}
if len(kc.token) > 0 {
req.Header.Set("Authorization", "Bearer "+kc.token)
}
if contentType != "" {
req.Header.Set("Content-Type", contentType)
}
return req, nil
}
func (kc *k8sClient) DeleteRequest(url string) (*http.Request, error) {
if !strings.HasPrefix(url, kc.host) {
url = fmt.Sprintf("%s/%s", kc.host, url)
}
req, err := http.NewRequest("DELETE", url, nil)
if err != nil {
return nil, err
}
if len(kc.token) > 0 {
req.Header.Set("Authorization", "Bearer "+kc.token)
}
return req, nil
}
func (kc *k8sClient) Do(req *http.Request) (*http.Response, error) {
return kc.httpClient.Do(req)
}
func (kc *k8sClient) Host() string {
return kc.host
}
// NewInClusterK8sClient creates K8sClient if it is inside Kubernetes
func NewInClusterK8sClient() (Client, error) {
host, port := os.Getenv("KUBERNETES_SERVICE_HOST"), os.Getenv("KUBERNETES_SERVICE_PORT")
if len(host) == 0 || len(port) == 0 {
return nil, fmt.Errorf("unable to load in-cluster configuration, KUBERNETES_SERVICE_HOST and KUBERNETES_SERVICE_PORT must be defined")
}
token, err := ioutil.ReadFile(serviceAccountToken)
if err != nil {
return nil, err
}
ca, err := ioutil.ReadFile(serviceAccountCACert)
if err != nil {
return nil, err
}
certPool := x509.NewCertPool()
certPool.AppendCertsFromPEM(ca)
transport := &http.Transport{TLSClientConfig: &tls.Config{
MinVersion: tls.VersionTLS10,
RootCAs: certPool,
}}
httpClient := &http.Client{Transport: transport, Timeout: time.Nanosecond * 0}
return &k8sClient{
host: "https://" + net.JoinHostPort(host, port),
token: string(token),
httpClient: httpClient,
}, nil
}
// NewInsecureK8sClient creates an insecure k8s client which is suitable
// to connect kubernetes api behind proxy
func NewInsecureK8sClient(apiURL string) Client {
return &k8sClient{
host: apiURL,
httpClient: http.DefaultClient,
}
}

@ -0,0 +1,601 @@
package main
import (
"context"
"crypto/sha256"
"encoding/json"
"encoding/hex"
"fmt"
"io/ioutil"
"net/http"
"os"
"strings"
"time"
"github.com/ghodss/yaml"
"github.com/pkg/errors"
log "github.com/sirupsen/logrus"
"github.com/smallstep/certificates/ca"
"github.com/smallstep/cli/crypto/pemutil"
"k8s.io/api/admission/v1beta1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
corev1 "k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/runtime"
"k8s.io/apimachinery/pkg/runtime/serializer"
)
var (
runtimeScheme = runtime.NewScheme()
codecs = serializer.NewCodecFactory(runtimeScheme)
deserializer = codecs.UniversalDeserializer()
// GetRootCAPath() is broken; points to secrets not certs. So
// we'll hard code instead for now.
//rootCAPath = pki.GetRootCAPath()
rootCAPath = "/home/step/.step/certs/root_ca.crt"
)
const (
admissionWebhookAnnotationKey = "autocert.step.sm/name"
admissionWebhookStatusKey = "autocert.step.sm/status"
provisionerPasswordFile = "/home/step/password/password"
volumeMountPath = "/var/run/autocert.step.sm"
tokenSecretKey = "token"
tokenSecretLabel = "autocert.step.sm/token"
)
// Config options for the autocert admission controller.
type Config struct {
LogFormat string `yaml:"logFormat"`
CaUrl string `yaml:"caUrl"`
CertLifetime string `yaml:"certLifetime"`
Bootstrapper corev1.Container `yaml:"bootstrapper"`
Renewer corev1.Container `yaml:"renewer"`
CertsVolume corev1.Volume `yaml:"certsVolume"`
}
// RFC6902 JSONPatch Operation
type PatchOperation struct {
Op string `json:"op"`
Path string `json:"path"`
Value interface{} `json:"value,omitempty"`
}
// RFC6901 JSONPath Escaping -- https://tools.ietf.org/html/rfc6901
func escapeJsonPath(path string) string {
// Replace`~` with `~0` then `/` with `~1`. Note that the order
// matters otherwise we'll turn a `/` into a `~/`.
path = strings.Replace(path, "~", "~0", -1)
path = strings.Replace(path, "/", "~1", -1)
return path
}
func loadConfig(file string) (*Config, error) {
data, err := ioutil.ReadFile(file)
if err != nil {
return nil, err
}
var cfg Config
if err := yaml.Unmarshal(data, &cfg); err != nil {
return nil, err
}
return &cfg, nil
}
// createTokenSecret generates a kubernetes Secret object containing a bootstrap token
// in the specified namespce. The secret name is randomly generated with a given prefix.
// A goroutine is scheduled to cleanup the secret after the token expires. The secret
// is also labelled for easy identification and manual cleanup.
func createTokenSecret(prefix, namespace, token string) (string, error) {
secret := corev1.Secret {
TypeMeta: metav1.TypeMeta {
Kind: "Secret",
APIVersion: "v1",
},
ObjectMeta: metav1.ObjectMeta {
GenerateName: prefix,
Namespace: namespace,
Labels: map[string]string {
tokenSecretLabel: "true",
},
},
StringData: map[string]string {
tokenSecretKey: token,
},
Type: corev1.SecretTypeOpaque,
}
client, err := NewInClusterK8sClient()
if err != nil {
return "", err
}
body, err := json.Marshal(secret)
if err != nil {
return "", err
}
log.WithField("secret", string(body)).Debug("Creating secret")
req, err := client.PostRequest(fmt.Sprintf("api/v1/namespaces/%s/secrets", namespace), string(body), "application/json")
if err != nil {
return "", err
}
resp, err := client.Do(req)
if err != nil {
log.Errorf("Secret creation error. Response: %v", resp)
return "", errors.Wrap(err, "secret creation")
}
if resp.StatusCode < 200 || resp.StatusCode >= 300 {
log.Errorf("Secret creation error (!2XX). Response: %v", resp)
var rbody []byte
if resp.Body != nil {
if data, err := ioutil.ReadAll(resp.Body); err == nil {
rbody = data
}
}
log.Error("Error body: ", string(rbody))
return "", errors.New("Not 200")
}
var rbody []byte
if resp.Body != nil {
if data, err := ioutil.ReadAll(resp.Body); err == nil {
rbody = data
}
}
if len(rbody) == 0 {
return "", errors.New("Empty response body")
}
var created *corev1.Secret
if err := json.Unmarshal(rbody, &created); err != nil {
return "", errors.Wrap(err, "Error unmarshalling secret response")
}
// Clean up after ourselves by deleting the Secret after the bootstrap
// token expires. This is best effort -- obviously we'll miss some stuff
// if this process goes away -- but the secrets are also labelled so
// it's also easy to clean them up in bulk using kubectl if we miss any.
go func() {
time.Sleep(tokenLifetime)
req, err := client.DeleteRequest(fmt.Sprintf("api/v1/namespaces/%s/secrets/%s", namespace, created.Name))
ctxLog := log.WithFields(log.Fields{
"name": created.Name,
"namespace": namespace,
})
if err != nil {
ctxLog.WithField("error", err).Error("Error deleting expired boostrap token secret")
return
}
resp, err := client.Do(req)
if err != nil {
ctxLog.WithField("error", err).Error("Error deleting expired boostrap token secret")
return
}
if resp.StatusCode < 200 || resp.StatusCode >= 300 {
ctxLog.WithFields(log.Fields{
"status": resp.Status,
"statusCode": resp.StatusCode,
}).Error("Error deleting expired boostrap token secret")
return
}
ctxLog.Info("Deleted expired bootstrap token secret")
}()
return created.Name, err
}
// mkBootstrapper generates a bootstrap container based on the template defined in Config. It
// generates a new bootstrap token and mounts it, along with other required coniguration, as
// environment variables in the returned bootstrap container.
func mkBootstrapper(config *Config, commonName string, namespace string, provisioner Provisioner) (corev1.Container, error) {
b := config.Bootstrapper
token, err := provisioner.Token(commonName)
if err != nil {
return b, errors.Wrap(err, "token generation")
}
// Generate CA fingerprint
crt, err := pemutil.ReadCertificate(rootCAPath)
if err != nil {
return b, errors.Wrap(err, "CA fingerprint")
}
sum := sha256.Sum256(crt.Raw)
fingerprint := strings.ToLower(hex.EncodeToString(sum[:]))
secretName, err := createTokenSecret(commonName + "-", namespace, token)
if err != nil {
return b, errors.Wrap(err, "create token secret")
}
log.Infof("Secret name is: %s", secretName)
b.Env = append(b.Env, corev1.EnvVar {
Name: "COMMON_NAME",
Value: commonName,
})
b.Env = append(b.Env, corev1.EnvVar {
Name: "STEP_TOKEN",
ValueFrom: &corev1.EnvVarSource {
SecretKeyRef: &corev1.SecretKeySelector {
LocalObjectReference: corev1.LocalObjectReference {
Name: secretName,
},
Key: tokenSecretKey,
},
},
})
b.Env = append(b.Env, corev1.EnvVar {
Name: "STEP_CA_URL",
Value: config.CaUrl,
})
b.Env = append(b.Env, corev1.EnvVar {
Name: "STEP_FINGERPRINT",
Value: fingerprint,
})
b.Env = append(b.Env, corev1.EnvVar {
Name: "STEP_NOT_AFTER",
Value: config.CertLifetime,
})
return b, nil
}
// mkRenewer generates a new renewer based on the template provided in Config.
func mkRenewer(config *Config) (corev1.Container) {
r := config.Renewer
r.Env = append(r.Env, corev1.EnvVar {
Name: "STEP_CA_URL",
Value: config.CaUrl,
})
return r
}
func addContainers(existing, new []corev1.Container, path string) (ops []PatchOperation) {
if len(existing) == 0 {
return []PatchOperation {
PatchOperation {
Op: "add",
Path: path,
Value: new,
},
}
} else {
for _, add := range new {
ops = append(ops, PatchOperation {
Op: "add",
Path: path + "/-",
Value: add,
})
}
return ops
}
}
func addVolumes(existing, new []corev1.Volume, path string) (ops []PatchOperation) {
if len(existing) == 0 {
return []PatchOperation{
PatchOperation {
Op: "add",
Path: path,
Value: new,
},
}
} else {
for _, add := range new {
ops = append(ops, PatchOperation {
Op: "add",
Path: path + "/-",
Value: add,
})
}
return ops
}
}
func addCertsVolumeMount(volumeName string, containers []corev1.Container) (ops []PatchOperation) {
volumeMount := corev1.VolumeMount {
Name: volumeName,
MountPath: volumeMountPath,
ReadOnly: true,
}
for i, container := range containers {
if len(container.VolumeMounts) == 0 {
ops = append(ops, PatchOperation {
Op: "add",
Path: fmt.Sprintf("/spec/containers/%v/volumeMounts", i),
Value: []corev1.VolumeMount{volumeMount},
})
} else {
ops = append(ops, PatchOperation {
Op: "add",
Path: fmt.Sprintf("/spec/containers/%v/volumeMounts/-", i),
Value: volumeMount,
})
}
}
return ops
}
func addAnnotations(existing, new map[string]string) (ops []PatchOperation) {
if len(existing) == 0 {
return []PatchOperation{
PatchOperation {
Op: "add",
Path: "/metadata/annotations",
Value: new,
},
}
}
for k, v := range new {
if existing[k] == "" {
ops = append(ops, PatchOperation {
Op: "add",
Path: "/metadata/annotations/" + escapeJsonPath(k),
Value: v,
})
} else {
ops = append(ops, PatchOperation {
Op: "replace",
Path: "/metadata/annotations/" + escapeJsonPath(k),
Value: v,
})
}
}
return ops
}
// patch produces a list of patches to apply to a pod to inject a certificate. In particular,
// we patch the pod in order to:
// - Mount the `certs` volume in existing containers defined in the pod
// - Add the autocert-renewer as a container (a sidecar)
// - Add the autocert-bootstrapper as an initContainer
// - Add the `certs` volume definition
// - Annotate the pod to indicate that it's been processed by this controller
// The result is a list of serialized JSONPatch objects (or an error).
func patch(pod *corev1.Pod, namespace string, config *Config, provisioner Provisioner) ([]byte, error) {
var ops[] PatchOperation
commonName := pod.ObjectMeta.GetAnnotations()[admissionWebhookAnnotationKey]
renewer := mkRenewer(config)
bootstrapper, err := mkBootstrapper(config, commonName, namespace, provisioner)
if err != nil {
return nil, err
}
ops = append(ops, addCertsVolumeMount(config.CertsVolume.Name, pod.Spec.Containers)...)
ops = append(ops, addContainers(pod.Spec.Containers, []corev1.Container{renewer}, "/spec/containers")...)
ops = append(ops, addContainers(pod.Spec.InitContainers, []corev1.Container{bootstrapper}, "/spec/initContainers")...)
ops = append(ops, addVolumes(pod.Spec.Volumes, []corev1.Volume{config.CertsVolume}, "/spec/volumes")...)
ops = append(ops, addAnnotations(pod.Annotations, map[string]string{admissionWebhookStatusKey: "injected"})...)
return json.Marshal(ops)
}
// shouldMutate checks whether a pod is subject to mutation by this admission controller. A pod
// is subject to mutation if it's annotated with the `admissionWebhookAnnotationKey` and if it
// has not already been processed (indicated by `admissionWebhookStatusKey` set to `injected`).
func shouldMutate(metadata *metav1.ObjectMeta) bool {
annotations := metadata.GetAnnotations()
if annotations == nil {
annotations = map[string]string{}
}
// Only mutate if the object is annotated appropriately (annotation key set) and we haven't
// mutated already (status key isn't set).
if annotations[admissionWebhookAnnotationKey] == "" || annotations[admissionWebhookStatusKey] == "injected" {
return false
} else {
return true
}
}
// mutate takes an `AdmissionReview`, determines whether it is subject to mutation, and returns
// an appropriate `AdmissionResponse` including patches or any errors that occurred.
func mutate(review *v1beta1.AdmissionReview, config *Config, provisioner Provisioner) *v1beta1.AdmissionResponse {
ctxLog := log.WithField("uid", review.Request.UID)
request := review.Request
var pod corev1.Pod
if err := json.Unmarshal(request.Object.Raw, &pod); err != nil {
ctxLog.WithField("error", err).Error("Error unmarshalling pod")
return &v1beta1.AdmissionResponse {
Allowed: false,
UID: request.UID,
Result: &metav1.Status {
Message: err.Error(),
},
}
}
ctxLog = ctxLog.WithFields(log.Fields{
"kind": request.Kind,
"operation": request.Operation,
"name": pod.Name,
"generateName": pod.GenerateName,
"namespace": request.Namespace,
"user": request.UserInfo,
})
if !shouldMutate(&pod.ObjectMeta) {
ctxLog.WithField("annotations", pod.Annotations).Info("Skipping mutation")
return &v1beta1.AdmissionResponse {
Allowed: true,
UID: request.UID,
}
}
patchBytes, err := patch(&pod, request.Namespace, config, provisioner)
if err != nil {
ctxLog.WithField("error", err).Error("Error generating patch")
return &v1beta1.AdmissionResponse {
Allowed: false,
UID: request.UID,
Result: &metav1.Status {
Message: err.Error(),
},
}
}
ctxLog.WithField("patch", string(patchBytes)).Info("Generated patch")
return &v1beta1.AdmissionResponse {
Allowed: true,
Patch: patchBytes,
UID: request.UID,
PatchType: func() *v1beta1.PatchType {
pt := v1beta1.PatchTypeJSONPatch
return &pt
}(),
}
}
func main() {
if len(os.Args) != 2 {
log.Errorf("Usage: %s <config>\n", os.Args[0])
os.Exit(1)
}
config, err := loadConfig(os.Args[1])
if err != nil {
panic(err)
}
log.SetOutput(os.Stdout)
if config.LogFormat == "json" {
log.SetFormatter(&log.JSONFormatter{})
}
if config.LogFormat == "text" {
log.SetFormatter(&log.TextFormatter{})
}
log.WithFields(log.Fields{
"config": config,
}).Info("Loaded config")
provisionerName := os.Getenv("PROVISIONER_NAME")
provisionerKid := os.Getenv("PROVISIONER_KID")
log.WithFields(log.Fields{
"provisionerName": provisionerName,
"provisionerKid": provisionerKid,
}).Info("Loaded provisioner configuration")
provisioner, err := NewProvisioner(provisionerName, provisionerKid, config.CaUrl, rootCAPath, provisionerPasswordFile)
if err != nil {
log.Errorf("Error loading provisioner: %v", err)
os.Exit(1)
}
log.WithFields(log.Fields{
"name": provisioner.Name(),
"kid": provisioner.Kid(),
}).Info("Loaded provisioner")
namespace := os.Getenv("NAMESPACE")
if namespace == "" {
log.Errorf("$NAMESPACE not set")
os.Exit(1)
}
name := fmt.Sprintf("autocert.%s.svc", namespace)
token, err := provisioner.Token(name)
if err != nil {
log.WithField("error", err).Errorf("Error generating bootstrap token during controller startup")
os.Exit(1)
}
log.WithField("name", name).Infof("Generated bootstrap token for controller")
// make sure to cancel the renew goroutine
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
srv, err := ca.BootstrapServer(ctx, token, &http.Server{
Addr: ":4443",
Handler: http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if r.URL.Path == "/healthz" {
log.Info("/healthz")
fmt.Fprintf(w, "ok")
w.WriteHeader(http.StatusOK)
return
}
/*
var name string
if r.TLS != nil && len(r.TLS.PeerCertificates) > 0 {
name = r.TLS.PeerCertificates[0].Subject.CommonName
}
*/
if r.URL.Path != "/mutate" {
log.WithField("path", r.URL.Path).Error("Bad Request: 404 Not Found")
http.NotFound(w, r)
return
}
var body []byte
if r.Body != nil {
if data, err := ioutil.ReadAll(r.Body); err == nil {
body = data
}
}
if len(body) == 0 {
log.Error("Bad Request: 400 (Empty Body)")
http.Error(w, "Bad Request (Empty Body)", http.StatusBadRequest)
return
}
contentType := r.Header.Get("Content-Type")
if contentType != "application/json" {
log.WithField("Content-Type", contentType).Error("Bad Request: 415 (Unsupported Media Type)")
http.Error(w, fmt.Sprintf("Bad Request: 415 Unsupported Media Type (Expected Content-Type 'application/json' but got '%s')", contentType), http.StatusUnsupportedMediaType)
return
}
var response *v1beta1.AdmissionResponse
review := v1beta1.AdmissionReview{}
if _, _, err := deserializer.Decode(body, nil, &review); err != nil {
log.WithFields(log.Fields{
"body": body,
"error": err,
}).Error("Can't decode body")
response = &v1beta1.AdmissionResponse {
Allowed: false,
Result: &metav1.Status {
Message: err.Error(),
},
}
} else {
response = mutate(&review, config, provisioner)
}
resp, err := json.Marshal(v1beta1.AdmissionReview {
Response: response,
})
if err != nil {
log.WithFields(log.Fields{
"uid": review.Request.UID,
"error": err,
}).Info("Marshal error")
http.Error(w, fmt.Sprintf("Marshal Error: %v", err), http.StatusInternalServerError)
} else {
log.WithFields(log.Fields{
"uid": review.Request.UID,
"response": string(resp),
}).Info("Returning review")
if _, err := w.Write(resp); err != nil {
log.WithFields(log.Fields{
"uid": review.Request.UID,
"error": err,
}).Info("Write error")
}
}
}),
}, ca.VerifyClientCertIfGiven())
if err != nil {
panic(err)
}
log.Info("Listening on :4443 ...")
if err := srv.ListenAndServeTLS("", ""); err != nil {
panic(err)
}
}

@ -0,0 +1,143 @@
package main
import (
"encoding/json"
"fmt"
"time"
"github.com/pkg/errors"
"github.com/smallstep/cli/crypto/pki"
"github.com/smallstep/cli/crypto/randutil"
"github.com/smallstep/cli/jose"
"github.com/smallstep/cli/token"
"github.com/smallstep/cli/token/provision"
)
const (
tokenLifetime = 5 * time.Minute
)
type Provisioner interface {
Name() string
Kid() string
Token(subject string) (string, error)
}
type provisioner struct {
name string
kid string
caUrl string
caRoot string
jwk *jose.JSONWebKey
tokenLifetime time.Duration
}
// Name returns the provisioner's name.
func (p *provisioner) Name() string {
return p.name
}
// Kid returns the provisioners key ID.
func (p *provisioner) Kid() string {
return p.kid
}
// Token generates a bootstrap token for a subject.
func (p *provisioner) Token(subject string) (string, error) {
// A random jwt id will be used to identify duplicated tokens
jwtID, err := randutil.Hex(64) // 256 bits
if err != nil {
return "", err
}
notBefore := time.Now()
notAfter := notBefore.Add(tokenLifetime)
signUrl := fmt.Sprintf("%v/1.0/sign", p.caUrl)
tokOptions := []token.Options{
token.WithJWTID(jwtID),
token.WithKid(p.kid),
token.WithIssuer(p.name),
token.WithAudience(signUrl),
token.WithValidity(notBefore, notAfter),
token.WithRootCA(p.caRoot),
}
tok, err := provision.New(subject, tokOptions...)
if err != nil {
return "", err
}
return tok.SignedString(p.jwk.Algorithm, p.jwk.Key)
}
func decryptProvisionerJWK(encryptedKey, passFile string) (*jose.JSONWebKey, error) {
decrypted, err := jose.Decrypt("", []byte(encryptedKey), jose.WithPasswordFile(passFile))
if err != nil {
return nil, err
}
jwk := new(jose.JSONWebKey)
if err := json.Unmarshal(decrypted, jwk); err != nil {
return nil, errors.Wrap(err, "error unmarshalling provisioning key")
}
return jwk, nil
}
// loadProvisionerJWKByKid retrieves a provisioner key from the CA by key ID and
// decrypts it using the specified password file.
func loadProvisionerJWKByKid(kid, caUrl, caRoot, passFile string) (*jose.JSONWebKey, error) {
encrypted, err := pki.GetProvisionerKey(caUrl, caRoot, kid)
if err != nil {
return nil, err
}
return decryptProvisionerJWK(encrypted, passFile)
}
// loadProvisionerJWKByName retrieves the list of provisioners and encrypted key then
// returns the key of the first provisioner with a matching name that can be successfully
// decrypted with the specified password file.
func loadProvisionerJWKByName(name, caUrl, caRoot, passFile string) (key *jose.JSONWebKey, err error) {
provisioners, err := pki.GetProvisioners(caUrl, caRoot)
if err != nil {
err = errors.Wrap(err, "error getting the provisioners")
return
}
for _, provisioner := range provisioners {
if provisioner.Name == name {
key, err = decryptProvisionerJWK(provisioner.EncryptedKey, passFile)
if err == nil {
return
}
}
}
return nil, errors.New(fmt.Sprintf("provisioner '%s' not found (or your password is wrong)", name))
}
// NewProvisioner loads and decrypts key material from the CA for the named
// provisioner. The key identified by `kid` will be used if specified. If `kid`
// is the empty string we'll use the first key for the named provisioner that
// decrypts using `passFile`.
func NewProvisioner(name, kid, caUrl, caRoot, passFile string) (Provisioner, error) {
var jwk *jose.JSONWebKey
var err error
if kid != "" {
jwk, err = loadProvisionerJWKByKid(kid, caUrl, caRoot, passFile)
} else {
jwk, err = loadProvisionerJWKByName(name, caUrl, caRoot, passFile)
}
if err != nil {
return nil, err
}
return &provisioner{
name: name,
kid: jwk.KeyID,
caUrl: caUrl,
caRoot: caRoot,
jwk: jwk,
tokenLifetime: tokenLifetime,
}, nil
}

Binary file not shown.

After

Width:  |  Height:  |  Size: 10 MiB

@ -0,0 +1,103 @@
# hello-mtls
This repository contains examples of dockerized [m]TLS clients and servers in
various languages. There's a lot of confusion and misinformation regarding how
to do mTLS properly with an internal public key infrastructure. The goal of
this repository is to demonstrate best practices like:
* Properly configuring TLS to use your internal CA's root certificate
* mTLS (client certificates / client authentication)
* Short-lived certificate support (clients and servers automatically load
renewed certificates)
Examples use multi-stage docker builds and can be built via without any
required local dependencies (except `docker`):
```
docker build -f Dockerfile.server -t hello-mtls-server-<lang> .
docker build -f Dockerfile.client -t hello-mtls-client-<lang> .
```
Once built, you should be able to deploy via:
```
kubectl apply -f hello-mtls.server.yaml
kubectl apply -f hello-mtls.client.yaml
```
## Mutual TLS
Unlike the _server auth TLS_ that's typical with web browsers, where the browser authenticates the server but not vice versa, _mutual TLS_ (mTLS) connections have both remote peers (client and server) authenticate to one another by presenting certificates. mTLS is not a different protocol. It's just a variant of TLS that's not usually turned on by default. This respository demonstrates **how to turn on mTLS** with different tools and languages. It also demonstrates other **TLS best practices** like certificate rotation.
mTLS provides _authenticated encryption_: an _identity dialtone_ and _end-to-end encryption_ for your workloads. It's like a secure line with caller ID. This has [all sorts of benefits](https://smallstep.com/blog/use-tls.html): better security, compliance, and easier auditability for starters. It **makes workloads identity-aware**, improving observability and enabling granular access control. Perhaps most compelling, mTLS lets you securely communicate with workloads running anywhere. Code, containers, devices, people, and anything else can connect securely using mTLS as long as they know one anothers' names and can resolve those names to routable IP addresses.
With properly configured mTLS, services can be safely exposed directly to the public internet: **only clients that have a certificate issued by the internal certificate authority will be allowed to connect**.
Here's a rough approximation of how an mTLS handshake works:
![mTLS handshake diagram](https://raw.githubusercontent.com/smallstep/certificates/autocert/autocert/mtls-handshake.png)
A few things to note:
* It's the signing of random numbers that proves we're talking to the right remote. It's the digital equivalent of asking someone to send you a photo of them with today's newspaper.
* The client and server need to have prior knowledge of the root certificate(s) used for signing other certificates.
* The client and server need to be configured to use the correct certificate and private key (the certificate must have been issued by a CA with a trusted root certificate)
* Private keys are never shared. This is the magic of public key cryptography: unlike passwords or access tokens, certificates let you prove who you are without giving anyone the ability to impersonate you.
## Feature matrix
This matrix shows the set of features we'd like to demonstrate in each language
and where each language is. Bug fixes, improvements, and examples in new
languages are appreciated!
[go/](go/)
- [X] Server using autocert certificate & key
- [X] mTLS (client authentication using internal root certificate)
- [X] Automatic certificate renewal
- [X] Restrict to safe ciphersuites and TLS versions
- [ ] TLS stack configuration loaded from `step-ca`
- [ ] Root certificate rotation
- [X] Client using autocert root certificate
- [X] mTLS (send client certificate if server asks for it)
- [ ] Automatic certificate rotation
- [X] Restrict to safe ciphersuites and TLS versions
- [ ] TLS stack configuration loaded from `step-ca`
- [ ] Root certificate rotation
[curl/](curl/)
- [X] Client
- [X] mTLS (send client certificate if server asks for it)
- [X] Automatic certificate rotation
- [ ] Restrict to safe ciphersuites and TLS versions
- [ ] TLS stack configuration loaded from `step-ca`
- [ ] Root certificate rotation
[nginx/](nginx/)
- [X] Server
- [X] mTLS (client authentication using internal root certificate)
- [X] Automatic certificate renewal
- [X] Restrict to safe ciphersuites and TLS versions
- [ ] TLS stack configuration loaded from `step-ca`
- [ ] Root certificate rotation
[node/](node/)
- [X] Server
- [X] mTLS (client authentication using internal root certificate)
- [X] Automatic certificate renewal
- [X] Restrict to safe ciphersuites and TLS versions
- [ ] TLS stack configuration loaded from `step-ca`
- [ ] Root certificate rotation
- [X] Client using autocert root certificate
- [X] mTLS (send client certificate if server asks for it)
- [X] Automatic certificate rotation
- [X] Restrict to safe ciphersuites and TLS versions
- [ ] TLS stack configuration loaded from `step-ca`
- [ ] Root certificate rotation
[envoy/](envoy/)
- [X] Server
- [X] mTLS (client authentication using internal root certificate)
- [X] Automatic certificate renewal
- [X] Restrict to safe ciphersuites and TLS versions
- [ ] TLS stack configuration loaded from `step-ca`
- [ ] Root certificate rotation

@ -0,0 +1,5 @@
FROM alpine
RUN apk add --no-cache bash curl
COPY client.sh .
RUN chmod +x client.sh
ENTRYPOINT ./client.sh

@ -0,0 +1,11 @@
#!/bin/bash
while :
do
response=$(curl -sS \
--cacert /var/run/autocert.step.sm/root.crt \
--cert /var/run/autocert.step.sm/site.crt \
--key /var/run/autocert.step.sm/site.key \
${HELLO_MTLS_URL})
echo "$(date): ${response}"
sleep 5
done

@ -0,0 +1,22 @@
apiVersion: apps/v1
kind: Deployment
metadata:
name: hello-mtls-client
labels: {app: hello-mtls-client}
spec:
replicas: 1
selector: {matchLabels: {app: hello-mtls-client}}
template:
metadata:
annotations:
autocert.step.sm/name: hello-mtls-client.default.pod.cluster.local
labels: {app: hello-mtls-client}
spec:
containers:
- name: hello-mtls-client
image: hello-mtls-client-curl:latest
imagePullPolicy: Never
resources: {requests: {cpu: 10m, memory: 20Mi}}
env:
- name: HELLO_MTLS_URL
value: https://hello-mtls.default.svc.cluster.local

@ -0,0 +1,21 @@
FROM envoyproxy/envoy-alpine
RUN apk update
RUN apk add python3
RUN apk add inotify-tools
RUN mkdir /src
ADD entrypoint.sh /src
ADD certwatch.sh /src
ADD hot-restarter.py /src
ADD start-envoy.sh /src
ADD server.yaml /src
# Flask app
ADD server.py /src
ADD requirements.txt /src
RUN pip3 install -r /src/requirements.txt
# app, certificate watcher and envoy
ENTRYPOINT ["/src/entrypoint.sh"]
CMD ["python3", "/src/hot-restarter.py", "/src/start-envoy.sh"]

@ -0,0 +1,6 @@
#!/bin/sh
while true; do
inotifywait -e modify /var/run/autocert.step.sm/site.crt
kill -HUP 1
done

@ -0,0 +1,10 @@
#!/bin/sh
# start hello world app
python3 /src/server.py &
# watch for the update of the cert and reload nginx
/src/certwatch.sh &
# Run docker CMD
exec "$@"

@ -0,0 +1,33 @@
apiVersion: v1
kind: Service
metadata:
labels: {app: hello-mtls}
name: hello-mtls
spec:
type: ClusterIP
ports:
- port: 443
targetPort: 443
selector: {app: hello-mtls}
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: hello-mtls
labels: {app: hello-mtls}
spec:
replicas: 1
selector: {matchLabels: {app: hello-mtls}}
template:
metadata:
annotations:
autocert.step.sm/name: hello-mtls.default.svc.cluster.local
labels: {app: hello-mtls}
spec:
containers:
- name: hello-mtls
image: hello-mtls-server-envoy:latest
imagePullPolicy: Never
resources: {requests: {cpu: 10m, memory: 20Mi}}

@ -0,0 +1,209 @@
#!/usr/bin/env python
from __future__ import print_function
import os
import signal
import sys
import time
# The number of seconds to wait for children to gracefully exit after
# propagating SIGTERM before force killing children.
# NOTE: If using a shutdown mechanism such as runit's `force-stop` which sends
# a KILL after a specified timeout period, it's important to ensure that this
# constant is smaller than the KILL timeout
TERM_WAIT_SECONDS = 30
restart_epoch = 0
pid_list = []
def term_all_children():
""" Iterate through all known child processes, send a TERM signal to each of
them, and then wait up to TERM_WAIT_SECONDS for them to exit gracefully,
exiting early if all children go away. If one or more children have not
exited after TERM_WAIT_SECONDS, they will be forcibly killed """
# First uninstall the SIGCHLD handler so that we don't get called again.
signal.signal(signal.SIGCHLD, signal.SIG_DFL)
global pid_list
for pid in pid_list:
print("sending TERM to PID={}".format(pid))
try:
os.kill(pid, signal.SIGTERM)
except OSError:
print("error sending TERM to PID={} continuing".format(pid))
all_exited = False
# wait for TERM_WAIT_SECONDS seconds for children to exit cleanly
retries = 0
while not all_exited and retries < TERM_WAIT_SECONDS:
for pid in list(pid_list):
ret_pid, exit_status = os.waitpid(pid, os.WNOHANG)
if ret_pid == 0 and exit_status == 0:
# the child is still running
continue
pid_list.remove(pid)
if len(pid_list) == 0:
all_exited = True
else:
retries += 1
time.sleep(1)
if all_exited:
print("all children exited cleanly")
else:
for pid in pid_list:
print("child PID={} did not exit cleanly, killing".format(pid))
force_kill_all_children()
sys.exit(1) # error status because a child did not exit cleanly
def force_kill_all_children():
""" Iterate through all known child processes and force kill them. Typically
term_all_children() should be attempted first to give child processes an
opportunity to clean up state before exiting """
global pid_list
for pid in pid_list:
print("force killing PID={}".format(pid))
try:
os.kill(pid, signal.SIGKILL)
except OSError:
print("error force killing PID={} continuing".format(pid))
pid_list = []
def shutdown():
""" Attempt to gracefully shutdown all child Envoy processes and then exit.
See term_all_children() for further discussion. """
term_all_children()
sys.exit(0)
def sigterm_handler(signum, frame):
""" Handler for SIGTERM. """
print("got SIGTERM")
shutdown()
def sigint_handler(signum, frame):
""" Handler for SIGINT (ctrl-c). The same as the SIGTERM handler. """
print("got SIGINT")
shutdown()
def sighup_handler(signum, frame):
""" Handler for SIGUP. This signal is used to cause the restarter to fork and exec a new
child. """
print("got SIGHUP")
fork_and_exec()
def sigusr1_handler(signum, frame):
""" Handler for SIGUSR1. Propagate SIGUSR1 to all of the child processes """
global pid_list
for pid in pid_list:
print("sending SIGUSR1 to PID={}".format(pid))
try:
os.kill(pid, signal.SIGUSR1)
except OSError:
print("error in SIGUSR1 to PID={} continuing".format(pid))
def sigchld_handler(signum, frame):
""" Handler for SIGCHLD. Iterates through all of our known child processes and figures out whether
the signal/exit was expected or not. Python doesn't have any of the native signal handlers
ability to get the child process info directly from the signal handler so we need to iterate
through all child processes and see what happened."""
print("got SIGCHLD")
kill_all_and_exit = False
global pid_list
pid_list_copy = list(pid_list)
for pid in pid_list_copy:
ret_pid, exit_status = os.waitpid(pid, os.WNOHANG)
if ret_pid == 0 and exit_status == 0:
# This child is still running.
continue
pid_list.remove(pid)
# Now we see how the child exited.
if os.WIFEXITED(exit_status):
exit_code = os.WEXITSTATUS(exit_status)
print("PID={} exited with code={}".format(ret_pid, exit_code))
if exit_code == 0:
# Normal exit. We assume this was on purpose.
pass
else:
# Something bad happened. We need to tear everything down so that whoever started the
# restarter can know about this situation and restart the whole thing.
kill_all_and_exit = True
elif os.WIFSIGNALED(exit_status):
print("PID={} was killed with signal={}".format(ret_pid, os.WTERMSIG(exit_status)))
kill_all_and_exit = True
else:
kill_all_and_exit = True
if kill_all_and_exit:
print("Due to abnormal exit, force killing all child processes and exiting")
# First uninstall the SIGCHLD handler so that we don't get called again.
signal.signal(signal.SIGCHLD, signal.SIG_DFL)
force_kill_all_children()
# Our last child died, so we have no purpose. Exit.
if not pid_list:
print("exiting due to lack of child processes")
sys.exit(1 if kill_all_and_exit else 0)
def fork_and_exec():
""" This routine forks and execs a new child process and keeps track of its PID. Before we fork,
set the current restart epoch in an env variable that processes can read if they care. """
global restart_epoch
os.environ['RESTART_EPOCH'] = str(restart_epoch)
print("forking and execing new child process at epoch {}".format(restart_epoch))
restart_epoch += 1
child_pid = os.fork()
if child_pid == 0:
# Child process
os.execl(sys.argv[1], sys.argv[1])
else:
# Parent process
print("forked new child process with PID={}".format(child_pid))
pid_list.append(child_pid)
def main():
""" Script main. This script is designed so that a process watcher like runit or monit can watch
this process and take corrective action if it ever goes away. """
print("starting hot-restarter with target: {}".format(sys.argv[1]))
signal.signal(signal.SIGTERM, sigterm_handler)
signal.signal(signal.SIGINT, sigint_handler)
signal.signal(signal.SIGHUP, sighup_handler)
signal.signal(signal.SIGCHLD, sigchld_handler)
signal.signal(signal.SIGUSR1, sigusr1_handler)
# Start the first child process and then go into an endless loop since everything else happens via
# signals.
fork_and_exec()
while True:
time.sleep(60)
if __name__ == '__main__':
main()

@ -0,0 +1,9 @@
from flask import Flask
app = Flask(__name__)
@app.route("/")
def hello():
return "Hello World!\n"
if __name__ == "__main__":
app.run(host='127.0.0.1', port=8080, debug=False)

@ -0,0 +1,50 @@
static_resources:
listeners:
- address:
socket_address:
address: 0.0.0.0
port_value: 443
filter_chains:
- filters:
- name: envoy.http_connection_manager
config:
codec_type: auto
stat_prefix: ingress_http
route_config:
name: hello
virtual_hosts:
- name: hello
domains:
- "hello-mtls.default.svc.cluster.local"
routes:
- match:
prefix: "/"
route:
cluster: hello-mTLS
http_filters:
- name: envoy.router
config: {}
tls_context:
common_tls_context:
tls_params:
tls_minimum_protocol_version: TLSv1_2
tls_maximum_protocol_version: TLSv1_3
cipher_suites: "[ECDHE-ECDSA-AES128-GCM-SHA256|ECDHE-ECDSA-CHACHA20-POLY1305]"
tls_certificates:
- certificate_chain:
filename: "/var/run/autocert.step.sm/site.crt"
private_key:
filename: "/var/run/autocert.step.sm/site.key"
validation_context:
trusted_ca:
filename: "/var/run/autocert.step.sm/root.crt"
require_client_certificate: true
clusters:
- name: hello-mTLS
connect_timeout: 0.25s
type: strict_dns
lb_policy: round_robin
hosts:
- socket_address:
address: 127.0.0.1
port_value: 8080

@ -0,0 +1,4 @@
#!/bin/sh
ulimit -n 65536
/usr/local/bin/envoy -c /src/server.yaml --service-cluster hello-mTLS --restart-epoch $RESTART_EPOCH

@ -0,0 +1,10 @@
# build stage
FROM golang:alpine AS build-env
RUN mkdir /src
ADD client.go /src
RUN cd /src && go build -o client
# final stage
FROM alpine
COPY --from=build-env /src/client .
ENTRYPOINT ./client

@ -0,0 +1,10 @@
# build stage
FROM golang:alpine AS build-env
RUN mkdir /src
ADD server.go /src
RUN cd /src && go build -o server
# final stage
FROM alpine
COPY --from=build-env /src/server .
ENTRYPOINT ./server

@ -0,0 +1,85 @@
package main
import (
"crypto/tls"
"crypto/x509"
"errors"
"fmt"
"io/ioutil"
"log"
"net/http"
"os"
"strings"
"time"
)
const (
autocertFile = "/var/run/autocert.step.sm/site.crt"
autocertKey = "/var/run/autocert.step.sm/site.key"
autocertRoot = "/var/run/autocert.step.sm/root.crt"
requestFrequency = 5 * time.Second
)
func loadRootCertPool() (*x509.CertPool, error) {
root, err := ioutil.ReadFile(autocertRoot)
if err != nil {
return nil, err
}
pool := x509.NewCertPool()
if ok := pool.AppendCertsFromPEM(root); !ok {
return nil, errors.New("Missing or invalid root certificate")
}
return pool, nil
}
func main() {
url := os.Getenv("HELLO_MTLS_URL")
// Read our leaf certificate and key from disk
cert, err := tls.LoadX509KeyPair(autocertFile, autocertKey)
if err != nil {
log.Fatal(err)
}
// Read the root certificate for our CA from disk
roots, err := loadRootCertPool()
if err != nil {
log.Fatal(err)
}
// Create an HTTPS client using our cert, key & pool
client := &http.Client{
Transport: &http.Transport{
TLSClientConfig: &tls.Config{
RootCAs: roots,
Certificates: []tls.Certificate{cert},
MinVersion: tls.VersionTLS12,
CurvePreferences: []tls.CurveID{tls.CurveP521, tls.CurveP384, tls.CurveP256},
CipherSuites: []uint16{
tls.TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305,
tls.TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256,
},
},
},
}
for {
// Make request
r, err := client.Get(url)
if err != nil {
log.Fatal(err)
}
defer r.Body.Close()
body, err := ioutil.ReadAll(r.Body)
if err != nil {
log.Fatal(err)
}
fmt.Printf("%s: %s\n", time.Now().Format(time.RFC3339), strings.Trim(string(body), "\n"))
time.Sleep(requestFrequency)
}
}

@ -0,0 +1,22 @@
apiVersion: apps/v1
kind: Deployment
metadata:
name: hello-mtls-client
labels: {app: hello-mtls-client}
spec:
replicas: 1
selector: {matchLabels: {app: hello-mtls-client}}
template:
metadata:
annotations:
autocert.step.sm/name: hello-mtls-client.default.pod.cluster.local
labels: {app: hello-mtls-client}
spec:
containers:
- name: hello-mtls-client
image: hello-mtls-client-go:latest
imagePullPolicy: Never
resources: {requests: {cpu: 10m, memory: 20Mi}}
env:
- name: HELLO_MTLS_URL
value: https://hello-mtls.default.svc.cluster.local

@ -0,0 +1,33 @@
apiVersion: v1
kind: Service
metadata:
labels: {app: hello-mtls}
name: hello-mtls
spec:
type: ClusterIP
ports:
- port: 443
targetPort: 443
selector: {app: hello-mtls}
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: hello-mtls
labels: {app: hello-mtls}
spec:
replicas: 1
selector: {matchLabels: {app: hello-mtls}}
template:
metadata:
annotations:
autocert.step.sm/name: hello-mtls.default.svc.cluster.local
labels: {app: hello-mtls}
spec:
containers:
- name: hello-mtls
image: hello-mtls-server-go:latest
imagePullPolicy: Never
resources: {requests: {cpu: 10m, memory: 20Mi}}

@ -0,0 +1,136 @@
package main
import (
"crypto/tls"
"crypto/x509"
"errors"
"fmt"
"io/ioutil"
"log"
"net/http"
"sync"
"time"
)
const (
autocertFile = "/var/run/autocert.step.sm/site.crt"
autocertKey = "/var/run/autocert.step.sm/site.key"
autocertRoot = "/var/run/autocert.step.sm/root.crt"
tickFrequency = 15 * time.Second
)
// Uses techniques from https://diogomonica.com/2017/01/11/hitless-tls-certificate-rotation-in-go/
// to automatically rotate certificates when they're renewed.
type rotator struct {
sync.Mutex
certificate *tls.Certificate
}
func (r *rotator) getCertificate(hello *tls.ClientHelloInfo) (*tls.Certificate, error) {
r.Lock()
defer r.Unlock()
return r.certificate, nil
}
func (r *rotator) loadCertificate(certFile, keyFile string) error {
r.Lock()
defer r.Unlock()
c, err := tls.LoadX509KeyPair(certFile, keyFile)
if err != nil {
return err
}
r.certificate = &c
return nil
}
func loadRootCertPool() (*x509.CertPool, error) {
root, err := ioutil.ReadFile(autocertRoot)
if err != nil {
return nil, err
}
pool := x509.NewCertPool()
if ok := pool.AppendCertsFromPEM(root); !ok {
return nil, errors.New("Missing or invalid root certificate")
}
return pool, nil
}
func main() {
mux := http.NewServeMux()
mux.HandleFunc("/", func(w http.ResponseWriter, r *http.Request) {
if r.TLS == nil || len(r.TLS.PeerCertificates) == 0 {
fmt.Fprintf(w, "Unauthenticated")
} else {
name := r.TLS.PeerCertificates[0].Subject.CommonName
fmt.Fprintf(w, "Hello, %s!\n", name)
}
})
mux.HandleFunc("/healthz", func(w http.ResponseWriter, r *http.Request) {
fmt.Fprintf(w, "Ok\n")
})
roots, err := loadRootCertPool()
if err != nil {
log.Fatal(err)
}
r := &rotator{}
cfg := &tls.Config{
ClientAuth: tls.RequireAndVerifyClientCert,
ClientCAs: roots,
MinVersion: tls.VersionTLS12,
CurvePreferences: []tls.CurveID{tls.CurveP521, tls.CurveP384, tls.CurveP256},
PreferServerCipherSuites: true,
CipherSuites: []uint16{
tls.TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305,
tls.TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256,
},
GetCertificate: r.getCertificate,
}
srv := &http.Server{
Addr: ":443",
Handler: mux,
TLSConfig: cfg,
}
// Load certificate
err = r.loadCertificate(autocertFile, autocertKey)
if err != nil {
log.Fatal("Error loading certificate and key", err)
}
// Schedule periodic re-load of certificate
done := make(chan struct{})
go func() {
ticker := time.NewTicker(tickFrequency)
defer ticker.Stop()
for {
select {
case <-ticker.C:
fmt.Println("Checking for new certificate...")
err := r.loadCertificate(autocertFile, autocertKey)
if err != nil {
log.Println("Error loading certificate and key", err)
}
case <- done:
return
}
}
}()
defer close(done)
log.Println("Listening no :443")
// Start serving HTTPS
err = srv.ListenAndServeTLS("", "")
if err != nil {
log.Fatal("ListenAndServerTLS: ", err)
}
}

@ -0,0 +1,11 @@
FROM nginx:alpine
RUN apk add inotify-tools
RUN mkdir /src
ADD site.conf /etc/nginx/conf.d
ADD certwatch.sh /src
ADD entrypoint.sh /src
# Certificate watcher and nginx
ENTRYPOINT ["/src/entrypoint.sh"]
CMD ["nginx", "-g", "daemon off;"]

@ -0,0 +1,6 @@
#!/bin/sh
while true; do
inotifywait -e modify /var/run/autocert.step.sm/site.crt
nginx -s reload
done

@ -0,0 +1,7 @@
#!/bin/sh
# watch for the update of the cert and reload nginx
/src/certwatch.sh &
# Run docker CMD
exec "$@"

@ -0,0 +1,33 @@
apiVersion: v1
kind: Service
metadata:
labels: {app: hello-mtls}
name: hello-mtls
spec:
type: ClusterIP
ports:
- port: 443
targetPort: 443
selector: {app: hello-mtls}
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: hello-mtls
labels: {app: hello-mtls}
spec:
replicas: 1
selector: {matchLabels: {app: hello-mtls}}
template:
metadata:
annotations:
autocert.step.sm/name: hello-mtls.default.svc.cluster.local
labels: {app: hello-mtls}
spec:
containers:
- name: hello-mtls
image: hello-mtls-server-nginx:latest
imagePullPolicy: Never
resources: {requests: {cpu: 10m, memory: 20Mi}}

@ -0,0 +1,16 @@
server {
listen 443 ssl;
server_name localhost;
ssl_protocols TLSv1.2;
ssl_certificate /var/run/autocert.step.sm/site.crt;
ssl_certificate_key /var/run/autocert.step.sm/site.key;
ssl_client_certificate /var/run/autocert.step.sm/root.crt;
ssl_verify_client on;
ssl_prefer_server_ciphers on;
ssl_ciphers ECDHE-ECDSA-CHACHA20-POLY1305:ECDHE-ECDSA-AES128-GCM-SHA256;
location / {
root /usr/share/nginx/html;
index index.html index.htm;
}
}

@ -0,0 +1,6 @@
FROM node:lts-alpine
RUN mkdir /src
ADD client.js /src
CMD ["node", "/src/client.js"]

@ -0,0 +1,6 @@
FROM node:lts-alpine
RUN mkdir /src
ADD server.js /src
CMD ["node", "/src/server.js"]

@ -0,0 +1,44 @@
const fs = require('fs');
const https = require('https');
const config = {
ca: '/var/run/autocert.step.sm/root.crt',
key: '/var/run/autocert.step.sm/site.key',
cert: '/var/run/autocert.step.sm/site.crt',
url: process.env.HELLO_MTLS_URL,
requestFrequency: 5000
};
var options = {
ca: fs.readFileSync(config.ca),
key: fs.readFileSync(config.key),
cert: fs.readFileSync(config.cert),
ciphers: 'ECDHE-ECDSA-CHACHA20-POLY1305:ECDHE-ECDSA-AES128-GCM-SHA256',
minVersion: 'TLSv1.2',
maxVersion: 'TLSv1.2',
// Not necessary as it defaults to true
rejectUnauthorized: true
};
fs.watch(config.cert, (event, filename) => {
if (event == 'change') {
options.cert = fs.readFileSync(config.cert);
}
});
function loop() {
var req = https.request(config.url, options, function(res) {
res.on('data', (data) => {
process.stdout.write(options.cert)
process.stdout.write(data)
setTimeout(loop, config.requestFrequency);
});
});
req.on('error', (e) => {
process.stderr.write('error: ' + e.message + '\n');
setTimeout(loop, config.requestFrequency);
})
req.end();
}
loop();

@ -0,0 +1,22 @@
apiVersion: apps/v1
kind: Deployment
metadata:
name: hello-mtls-client
labels: {app: hello-mtls-client}
spec:
replicas: 1
selector: {matchLabels: {app: hello-mtls-client}}
template:
metadata:
annotations:
autocert.step.sm/name: hello-mtls-client.default.pod.cluster.local
labels: {app: hello-mtls-client}
spec:
containers:
- name: hello-mtls-client
image: hello-mtls-client-node:latest
imagePullPolicy: Never
resources: {requests: {cpu: 10m, memory: 20Mi}}
env:
- name: HELLO_MTLS_URL
value: https://hello-mtls.default.svc.cluster.local

@ -0,0 +1,33 @@
apiVersion: v1
kind: Service
metadata:
labels: {app: hello-mtls}
name: hello-mtls
spec:
type: ClusterIP
ports:
- port: 443
targetPort: 443
selector: {app: hello-mtls}
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: hello-mtls
labels: {app: hello-mtls}
spec:
replicas: 1
selector: {matchLabels: {app: hello-mtls}}
template:
metadata:
annotations:
autocert.step.sm/name: hello-mtls.default.svc.cluster.local
labels: {app: hello-mtls}
spec:
containers:
- name: hello-mtls
image: hello-mtls-server-node:latest
imagePullPolicy: Never
resources: {requests: {cpu: 10m, memory: 20Mi}}

@ -0,0 +1,42 @@
const https = require('https');
const tls = require('tls');
const fs = require('fs');
var config = {
ca: '/var/run/autocert.step.sm/root.crt',
key: '/var/run/autocert.step.sm/site.key',
cert: '/var/run/autocert.step.sm/site.crt',
ciphers: 'ECDHE-ECDSA-CHACHA20-POLY1305:ECDHE-ECDSA-AES128-GCM-SHA256',
minVersion: 'TLSv1.2',
maxVersion: 'TLSv1.2'
};
function createSecureContext() {
return tls.createSecureContext({
ca: fs.readFileSync(config.ca),
key: fs.readFileSync(config.key),
cert: fs.readFileSync(config.cert),
ciphers: config.ciphers,
});
}
var ctx = createSecureContext()
fs.watch(config.cert, (event, filename) => {
if (event == 'change') {
ctx = createSecureContext();
}
});
https.createServer({
requestCert: true,
rejectUnauthorized: true,
SNICallback: (servername, cb) => {
cb(null, ctx);
}
}, (req, res) => {
res.writeHead(200);
res.end('hello nodejs\n');
}).listen(443);
console.log("Listening on :443 ...");

@ -0,0 +1,18 @@
FROM smallstep/step-cli:0.8.4-rc.1
ENV CA_NAME="Autocert"
ENV CA_DNS="ca.step.svc.cluster.local,127.0.0.1"
ENV CA_ADDRESS=":4443"
ENV CA_DEFAULT_PROVISIONER="admin"
ENV CA_URL="ca.step.svc.cluster.local"
ENV KUBE_LATEST_VERSION="v1.13.2"
USER root
RUN curl -L https://storage.googleapis.com/kubernetes-release/release/${KUBE_LATEST_VERSION}/bin/linux/amd64/kubectl -o /usr/local/bin/kubectl \
&& chmod +x /usr/local/bin/kubectl
RUN apk --update add expect
COPY autocert.sh /home/step/
RUN chmod +x /home/step/autocert.sh
CMD ["/home/step/autocert.sh"]

@ -0,0 +1,164 @@
#!/bin/bash
#set -x
echo "Welcome to Autocert configuration. Press return to begin."
read ANYKEY
STEPPATH=/home/step/.step
CA_PASSWORD=$(head /dev/urandom | tr -dc A-Za-z0-9 | head -c 32 ; echo '')
AUTOCERT_PASSWORD=$(head /dev/urandom | tr -dc A-Za-z0-9 | head -c 32 ; echo '')
echo -e "\e[1mChecking cluster permissions...\e[0m"
function permission_error {
# TODO: Figure out the actual service account instead of assuming default.
echo
echo -e "\033[0;31mPERMISSION ERROR\033[0m"
echo "Set permissions by running the following command, then try again:"
echo -e "\e[1m"
echo " kubectl create clusterrolebinding autocert-init-binding \\"
echo " --clusterrole cluster-admin \\"
echo " --user \"system:serviceaccount:default:default\""
echo -e "\e[0m"
echo "Once setup is complete you can remove this binding by running:"
echo -e "\e[1m"
echo " kubectl delete clusterrolebinding autocert-init-binding"
echo -e "\e[0m"
exit 1
}
echo -n "Checking for permission to create step namespace: "
kubectl auth can-i create namespaces
if [ $? -ne 0 ]; then
permission_error "create step namespace"
fi
echo -n "Checking for permission to create configmaps in step namespace: "
kubectl auth can-i create configmaps --namespace step
if [ $? -ne 0 ]; then
permission_error "create configmaps"
fi
echo -n "Checking for permission to create secrets in step namespace: "
kubectl auth can-i create secrets --namespace step
if [ $? -ne 0 ]; then
permission_error "create secrets"
fi
echo -n "Checking for permission to create deployments in step namespace: "
kubectl auth can-i create deployments --namespace step
if [ $? -ne 0 ]; then
permission_error "create deployments"
fi
echo -n "Checking for permission to create services in step namespace: "
kubectl auth can-i create services --namespace step
if [ $? -ne 0 ]; then
permission_error "create services"
fi
echo -n "Checking for permission to create cluster role: "
kubectl auth can-i create clusterrole
if [ $? -ne 0 ]; then
permission_error "create cluster roles"
fi
echo -n "Checking for permission to create cluster role binding:"
kubectl auth can-i create clusterrolebinding
if [ $? -ne 0 ]; then
permission_error "create cluster role bindings"
exit 1
fi
# Setting this here on purpose, after the above section which explicitly checks
# for and handles exit errors.
set -e
step ca init \
--name "$CA_NAME" \
--dns "$CA_DNS" \
--address "$CA_ADDRESS" \
--provisioner "$CA_DEFAULT_PROVISIONER" \
--with-ca-url "$CA_URL" \
--password-file <(echo "$CA_PASSWORD")
echo
echo -e "\e[1mCreating autocert provisioner...\e[0m"
expect <<EOD
spawn step ca provisioner add autocert --create
expect "Please enter a password to encrypt the provisioner private key? \\\\\\[leave empty and we'll generate one\\\\\\]: "
send "${AUTOCERT_PASSWORD}\n"
expect eof
EOD
echo
echo -e "\e[1mCreating step namespace and preparing environment...\e[0m"
kubectl create namespace step
kubectl -n step create configmap config --from-file $(step path)/config
kubectl -n step create configmap certs --from-file $(step path)/certs
kubectl -n step create configmap secrets --from-file $(step path)/secrets
kubectl -n step create secret generic ca-password --from-literal "password=${CA_PASSWORD}"
kubectl -n step create secret generic autocert-password --from-literal "password=${AUTOCERT_PASSWORD}"
# Deploy CA and wait for rollout to complete
echo
echo -e "\e[1mDeploying certificate authority...\e[0m"
kubectl apply -f https://raw.githubusercontent.com/smallstep/certificates/autocert/autocert/install/01-step-ca.yaml
kubectl -n step rollout status deployment/ca
# Deploy autocert, setup RBAC, and wait for rollout to complete
echo
echo -e "\e[1mDeploying autocert...\e[0m"
kubectl apply -f https://raw.githubusercontent.com/smallstep/certificates/autocert/autocert/install/02-autocert.yaml
kubectl apply -f https://raw.githubusercontent.com/smallstep/certificates/autocert/autocert/install/03-rbac.yaml
kubectl -n step rollout status deployment/autocert
# Some `base64`s wrap lines... no thanks!
CA_BUNDLE=$(cat $(step path)/certs/root_ca.crt | base64 | tr -d '\n')
cat <<EOF | kubectl apply -f -
apiVersion: admissionregistration.k8s.io/v1beta1
kind: MutatingWebhookConfiguration
metadata:
name: autocert-webhook-config
labels: {app: autocert}
webhooks:
- name: autocert.step.sm
clientConfig:
service:
name: autocert
namespace: step
path: "/mutate"
caBundle: $CA_BUNDLE
rules:
- operations: ["CREATE"]
apiGroups: [""]
apiVersions: ["v1"]
resources: ["pods"]
failurePolicy: Ignore
namespaceSelector:
matchLabels:
autocert.step.sm: enabled
EOF
FINGERPRINT=$(step certificate fingerprint $(step path)/certs/root_ca.crt)
echo
echo -e "\e[1mAutocert installed!\e[0m"
echo
echo "Store this information somewhere safe:"
echo " CA & admin provisioner password: ${CA_PASSWORD}"
echo " Autocert password: ${AUTOCERT_PASSWORD}"
echo " CA Fingerprint: ${FINGERPRINT}"
echo

@ -0,0 +1,87 @@
apiVersion: v1
kind: Service
metadata:
labels:
app: ca
name: ca
namespace: step
spec:
type: ClusterIP
ports:
- port: 443
targetPort: 4443
selector:
app: ca
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: ca
namespace: step
labels:
app: ca
spec:
replicas: 1
selector:
matchLabels:
app: ca
template:
metadata:
labels:
app: ca
spec:
containers:
- name: ca
image: smallstep/step-ca:0.8.3
env:
- name: PWDPATH
value: /home/step/password/password
resources:
requests:
cpu: 100m
memory: 20Mi
readinessProbe:
httpGet:
path: /health
port: 4443
scheme: HTTPS
initialDelaySeconds: 3
periodSeconds: 3
livenessProbe:
httpGet:
path: /health
port: 4443
scheme: HTTPS
initialDelaySeconds: 3
periodSeconds: 3
volumeMounts:
- name: config
mountPath: /home/step/.step/config
readOnly: true
- name: certs
mountPath: /home/step/.step/certs
readOnly: true
- name: secrets
mountPath: /home/step/.step/secrets
readOnly: true
- name: ca-password
mountPath: /home/step/password
readOnly: true
securityContext:
runAsUser: 1000
allowPrivilegeEscalation: false
volumes:
- name: certs
configMap:
name: certs
- name: config
configMap:
name: config
- name: secrets
configMap:
name: secrets
- name: ca-password
secret:
secretName: ca-password

@ -0,0 +1,106 @@
apiVersion: v1
kind: Service
metadata:
labels: {app: autocert}
name: autocert
namespace: step
spec:
type: ClusterIP
ports:
- port: 443
targetPort: 4443
selector: {app: autocert}
---
apiVersion: v1
kind: ConfigMap
metadata:
name: autocert-config
namespace: step
data:
config.yaml: |
logFormat: json # or text
caUrl: https://ca.step.svc.cluster.local
certLifetime: 24h
renewer:
name: autocert-renewer
image: smallstep/autocert-renewer:0.8.3
resources: {requests: {cpu: 10m, memory: 20Mi}}
imagePullPolicy: IfNotPresent
volumeMounts:
- name: certs
mountPath: /var/run/autocert.step.sm
bootstrapper:
name: autocert-bootstrapper
image: smallstep/autocert-bootstrapper:0.8.3
resources: {requests: {cpu: 10m, memory: 20Mi}}
imagePullPolicy: IfNotPresent
volumeMounts:
- name: certs
mountPath: /var/run/autocert.step.sm
certsVolume:
name: certs
emptyDir: {}
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: autocert
namespace: step
labels: {app: autocert}
spec:
replicas: 1
selector: {matchLabels: {app: autocert}}
template:
metadata: {labels: {app: autocert}}
spec:
containers:
- name: autocert
image: smallstep/autocert-controller:0.8.3
resources: {requests: {cpu: 100m, memory: 20Mi}}
env:
- name: PROVISIONER_NAME
value: autocert
- name: NAMESPACE
valueFrom:
fieldRef:
fieldPath: metadata.namespace
volumeMounts:
- name: config
mountPath: /home/step/.step/config
readOnly: true
- name: certs
mountPath: /home/step/.step/certs
readOnly: true
- name: autocert-password
mountPath: /home/step/password
readOnly: true
- name: autocert-config
mountPath: /home/step/autocert
readOnly: true
securityContext:
runAsUser: 1000
allowPrivilegeEscalation: false
livenessProbe:
httpGet:
path: /healthz
port: 4443
scheme: HTTPS
readinessProbe:
httpGet:
path: /healthz
port: 4443
scheme: HTTPS
volumes:
- name: config
configMap: {name: config}
- name: certs
configMap: {name: certs}
- name: autocert-password
secret: {secretName: autocert-password}
- name: autocert-config
configMap: {name: autocert-config}

@ -0,0 +1,36 @@
# Create a ClusterRole for managing autocert secrets, which should
# only exist in namespaces with autocert enabled and should always
# be labeled `autocert.step.sm/token: true`.
#
# To create this ClusterRole you need cluster-admin privileges. On
# GKE you can give yourself cluster-admin privileges using the
# following command:
#
# kubectl create clusterrolebinding cluster-admin-binding \
# --clusterrole cluster-admin \
# --user $(gcloud config get-value account)
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
name: autocert-controller
rules:
- apiGroups: [""]
resources: ["secrets"]
verbs: ["create", "delete"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
name: autocert-controller
roleRef:
apiGroup: rbac.authorization.k8s.io
kind: ClusterRole
name: autocert-controller
subjects:
- kind: ServiceAccount
name: default
namespace: step

Binary file not shown.

After

Width:  |  Height:  |  Size: 1.3 MiB

@ -0,0 +1,8 @@
FROM smallstep/step-cli:0.8.3
USER root
ENV CRT="/var/run/autocert.step.sm/site.crt"
ENV KEY="/var/run/autocert.step.sm/site.key"
ENV STEP_ROOT="/var/run/autocert.step.sm/root.crt"
ENTRYPOINT ["/bin/bash", "-c", "step ca renew --daemon $CRT $KEY"]
Loading…
Cancel
Save