✅Prerequisites
Hardware requirements
Minimum requirements for Toucan are:
An x86 Linux-based server. No requirements on the distribution or Kernel version.
2 Cores CPU
10 GB of RAM
20 GB of storage for the node, to host Kubernetes and the container images.
+Additional block storage for your data (at least 50 GB).
Benchmarks
These results show the average memory usage of an initial deployment. This is not the average memory usage of the stack under load.
keycloak 587Mi
dataexecution-api 18Mi
dataexecution-dragonfly 13Mi
dataexecution-worker 25Mi
dataset 585Mi
garage-0 103Mi
gotenberg 12Mi
impersonate 20Mi
impersonate-dragonfly 21Mi
laputa-0 2171Mi
laputa-dragonfly 21Mi
layout 217Mi
layout-dragonfly 19Mi
mongodb-0 225Mi
nats-0 16Mi
postgresql-0 136Mi
spicedb 98Mi
tucana 26Mi
vault-server-0 178MiRecommended requirements for Toucan are:
An x86 Linux-based server. No requirements on the distribution or Kernel version.
4 Cores CPU
16 GB of RAM
>50 GB of storage for the node.
+Additional block storage for your data (at least 50 GB).
Software requirements
Kubectl with an access to the Kubernetes cluster.
An access to Toucan's Quay registry. If you don't have access, contact us using our mail address: go@toucantoco.com.
Test the access by running the following command:
A configurable DNS to forward traffic to the Toucan services.
An Ingress controller like the Nginx Ingress controller to expose the Toucan services.
OR a Gateway controller like the Envoy Gateway to expose the Toucan services.
(optional) cert-manager to handle certificates.
Prerequisite knowledge
Even if the deployment is abstracted, basic understanding of components is recommended:
Kubernetes critical concepts: Gateway API (or Ingress) and Persistent Volumes.
Envoy Gateway (or NGINX Ingress controller) and reverse proxies
Required to expose traffic to the Toucan services.
DNS records configuration (feel free to check the documentation of your DNS cloud provider, here's the one provided by Cloudflare).
TLS configuration
Required to secure traffic to the Toucan services.
(optional) cert-manager to automatically manage certificates.
Are you self-hosting Kubernetes? Please read this.
If this is your first time configuring Kubernetes, exposing traffic through the Ingress Controller may not be immediate. We recommend using one of the following methods in order of preference:
Cloud Controller Manager (if possible): This is the optimal choice for multi-node deployments, as it automatically handles network configuration without the hassle of manual setup.
MetalLB: A solution for exposing traffic by announcing IP addresses of the Ingress Controller using L2 or BGP advertisements. MetalLB is ideal for multi-node setups where you need to manage load balancing and expose services over external IPs.
NodePorts: This method opens ports between 30000 and 32767 on each node, but you’ll need to manually forward ports using your router. This is a simple solution but may not scale well in larger environments.
Storage may also require additional configuration, and we recommend the following methods:
Cloud Provider Storage Provisioner: Using your cloud provider’s storage solution (e.g., AWS EBS, Google Persistent Disk, or Azure Disk) provides the benefits of an optimized storage plane, including replication, availability, and compliance. This is the most robust option for production-grade storage.
Local-Path Provisioner: Included with k3s, this provisioner mounts storage directly on the host. It's highly recommended for single-node setups but has the limitation of binding the Pod to a specific node. To optimize this setup, we suggest mounting an additional block storage at
/opt/local-path-provisioner/to separate application data from infrastructure data on the same disk.Other CSI Drivers (e.g., iSCSI): If you prefer, you can use other Container Storage Interface (CSI) drivers, such as iSCSI, for more custom storage configurations.
We do not recommend using self-hosted solutions like Ceph/Rook, NFS, Longhorn, or any other parallel/network storage unless you have dedicated storage nodes and a deep understanding of inter-node communication and performance tuning. These solutions typically require significant network bandwidth and can complicate setup and maintenance.
If you ever have storage nodes, for a multi node setup, we recommend splitting the storage between the compute nodes. You can then apply taints to designate certain nodes as "storage nodes" (e.g., storage=true), optimizing your storage layout:
If this is your first time self-hosting Kubernetes, we strongly recommend using k3s as your Kubernetes distribution. It simplifies network and storage configuration, making it the easiest way to get your cluster running with minimal effort.
Recommended tools
We recommend using the following tools:
While kubectl should be enough to interact with the cluster, these additional tools will help you quickly navigate through the cluster.
Recommended documentation
Warning
Before deploying Toucan Stack, it is strongly advised to have a comprehensive understanding on how Kubernetes operates, specifically with regard to storage and network management using features such as PersistentVolume, and Ingress.
What's next
Did you get everything ready?
Now it's time to deploy the stack! We'll also help you configure Kubernetes with the k3s distribution, if you haven't already deployed it.
⛴️Deploy Toucan using Helm ChartsLast updated
Was this helpful?