Using Kubernetes

Deploy Ayakaleaf Pro on a Kubernetes cluster with plain manifests.

The Toolkit is the recommended way to run Ayakaleaf Pro on a single server. If you already run a Kubernetes cluster, this section deploys the same container on it with plain kubectl manifests: no extra image, no Helm chart.

By the end of this section you have one Ayakaleaf Pro instance with sandboxed compiles, a MongoDB replica set and Redis, all in the namespace overleaf. To run several instances behind a load balancer, continue with Horizontal scaling afterwards.

Kubernetes support is experimental. We don’t publish a Helm chart or Kubernetes-specific images. The manifests in these pages are the ones we tested on Kubernetes 1.37 (kubeadm, containerd 2.4, flannel) with three nodes of 2 vCPUs and 4 GB of memory each. Adapt storage, networking and ingress to your cluster.

How the pieces fit

Toolkit (one server)	Kubernetes
config/variables.env	ConfigMap overleaf-env and Secret overleaf-secrets
sharelatex container	Deployment overleaf (one pod running all Overleaf services)
mongo and redis containers	StatefulSets mongo (replica set) and redis, or your managed services
~/overleaf-toolkit/data/overleaf	PersistentVolumeClaim overleaf-data
Docker on the host, used for compiles	Docker Engine on every node that runs Ayakaleaf Pro
bin/up, bin/logs, bin/shell	kubectl apply, kubectl logs, kubectl exec

What you need

• A Kubernetes cluster you can install packages on: every node that may run Ayakaleaf Pro needs Docker Engine, because compiles run in TeX Live containers that Ayakaleaf Pro starts through the node’s /var/run/docker.sock. See 0. Prepare the nodes.
• A default StorageClass with ReadWriteOnce volumes.
• kubectl and openssl on your workstation.