I recently started a Cardano Stake Pool, [CODER], hosted using Kubernetes. This is the first of several posts on how I set it up. The code/config in this post will be taken from my mainnet configuration, though I would recommend starting with testnet when starting out.
Posts in this series:
Step 1: Setting up Cardano Relays using Kubernetes/microk8s
Step 2: Monitoring Cardano Relays on Kubernetes with Grafana and Prometheus
Goals
My goals were:
Use Kubernetes
Part of the reason for this project was to have a real project to learn more about Kubernetes
I’d like to manage the nodes without needing to set up individual machines or manage individual containers via Docker
I’d like to be able to migrate/rebuild the pool (eg. from Intel NUCs to the cloud) later with as little effort as possible
Use Grafana/Prometheus to monitor the nodes without having to learn too much about them or set them up myself (specifically via the microk8s addons and ServiceMonitors)
Secure the pool pledge and rewards with a crypto hardware wallet so if any node was compromised or destroyed, none of these are lost
This is my first time using Kubernetes so what’s described here may not be the optimal solution. If you do have suggestions/improvements, please do comment below!
If you find this post useful or are looking for somewhere to delegate while setting up your own pool, check out my pool [CODER] which donates 30% of rewards to charities that help people into coding �
Installing Kubernetes (microk8s)
The first step is setting up Kubernetes. I decided to go with microk8s by Canonical since it seemed to have as good reviews as any and I figured being from Canonical there would be fewer potential compatibility issues with Ubuntu Server. It also included addons for Prometheus/Grafana which looked like it may simplify setting those up. Installing it turned out to be rather trivial, as the Ubuntu Server installation wizard offered it as an option in the final step (which I believe just installs the Snap package).
Node Volumes
By default, changes made to a containers file system will be lost when the container is terminated/restarted. Since it takes many hours to sync the Cardano blockchain, I knew I’d need some way to persist its db folder. For simplicity, I decided (for now) to just map some folders from the host machine into the containers. While I only have one Kubernetes node right now, this might not always be the case so the volumes need to be configured to be specific to this node (and if I decide to balance the relays across nodes, I would manually copy the data and then update the nodeAffinity section).
Again, I created a .yml file to hold the config and applied it with microk8s.kubectl apply -f ...yml:
Node Configuration Files
To run a node we’ll need some config files. For mainnet they can be downloaded like this:
I renamed the topology.json file to relay-topology.json since we’ll eventually also have a producer-topology.json that will be different, and I want to be able to collect/back up the files in one place.
StatefulSet/Pod Definition
StatefulSets in Kubernetes are like Deployments but with sticky identities. Rather than pod names being suffixed with some random key that changes upon destroy/recreate, the pods in a StatefulSet will be numbered, with the identities reused if a pod is destroyed/recreated. This helps if you need other services to know identities of each pod but also simplifies using the pod names in kubectl commands without having to keep looking them up if you’re constantly recreating them �
Again, I created a .yml file and microk8s.kubectl apply -f‘d it.
Exposing the Relay with NodePort
Next, we need to ensure the node is accessible to the world. We need inbound peers to be able to get mempool transactions (which we’ll ultimately need to include in any blocks produced by our producer). This uses a NodePort, just like the Grafana service above. One unresolved niggle I have is that this exposes only a single port and balances across multiple relays. This means right now I can only provide other peers a single relay hostname/port even though in reality I may be running multiple (if anyone has a solution to this, I’d love to know!).
With this applied, I shared my relay details with a few friendly SPOs and used the log files to verify I had incoming connections and transactions. Step one, complete!
If you find this post useful or are looking for somewhere to delegate while setting up your own pool, check out my pool [CODER] which donates 30% of rewards to charities that help people into coding �
This post was served up via my RSS feed. Please visit the original article to read/post comments. If you found this article interesting, why not follow @DanTup on Twitter for more? :)