/jitsi-kubernetes

Jitsi deployment on Kubernetes with JVB autoscale and OCTO region enabled

MIT LicenseMIT

jitsi kubernetes scalable service with octo

Jitsi Kubernetes deployment with autoscale JVB and OCTO enabled

Kubernetes 1.18.10

This repo is based on https://github.com/DushmanthaBandaranayake/jitsi-kubernetes-scalable-service

Thanks for this great work but I need to make it a different way to enable multi shards and regional with octo.

Prequirements

You need 2 kubernetes for each region, 1 for Main Jitsi Web Prosody and 1 for JVBs. If you have 2 regions, 4 kubernertes needed.

This separation make sure JVB autoscale not disturbe the Main Jitis Web as I see on Digitalocean kubernetes.

JVB nodes need at least 2 cpu (recommended 4).

Installation

0. Search for all places in the code marked as: << update this >> and update them!

1. Deploy Main Jitsi Web server on kubernetes shard 0:

Connect kubectl to kuberenets shard 0

Create kubernetes namespace

kubectl create namespace jitsi

Go to /base

kubectl apply -f config.yaml

Go to /shard0/web

kubectl apply -f jicofo.yaml

kubectl apply -f web-configmap.yaml

Go to /base/web-base

kubectl apply -f web-configmap.yaml

kubectl apply -f jicofo-configmap.yaml

kubectl apply -f service.yaml

kubectl apply -f web-prosody.yaml

2. Deploy the JVBs on kubernetes jvb-shard0

Connect kubectl to kuberenets jvb-shard0

Make sure udp port open on jvb node: udp 31000-30006 and OCTO udp port 30960 - 30966. with Digitalocean can use firewall with tag start with k8.. added by default to apply firewall to all nodes.

kubectl create namespace jitsi

Go to /base

kubectl apply -f config.yaml

Go to /jvb-base

kubectl apply -f server_metrics.yaml

kubectl apply -f jvb-configmap.yaml

kubectl apply -f service.yaml

Go to /shard0/jvb

kubectl apply -f jvb-statefullset.yaml

Your Jitsi meet now already available on first region with load balancing JVBs autoscale. Follow next steps to add more region if you have.

3. Deploy the second Jitsi Web server region on kubernetes shard1 (optional)

Connect kubectl to kuberenets shard1

Create kubernetes namespace

kubectl create namespace jitsi

Go to /base

kubectl apply -f config.yaml

Go to /shard1/web

kubectl apply -f web-configmap.yaml

Go to /base/web-base

kubectl apply -f web-configmap.yaml

kubectl apply -f service.yaml

kubectl apply -f web-prosody.yaml

4. Deploy the second JVBs on kubernetes jvb-shard1 for second region (if you have step 3)

Connect kubectl to kuberenets jvb-shard1

Make sure udp port open on jvb node: udp 31000-30006 and OCTO udp port 30960 - 30966

kubectl create namespace jitsi

Go to /base

kubectl apply -f config.yaml

Go to /jvb-base

kubectl apply -f server_metrics.yaml

kubectl apply -f jvb-configmap.yaml

kubectl apply -f service.yaml

Go to /shard1/jvb

kubectl apply -f jvb-statefullset.yaml

Your scaleable jitsi with octo will be avaiable at the main domain!

On Digitalocean point your domain to the load balancer created on kuberenets main web and second region one.

You can use route53 with regional routing support, this will select best domain for user. Just point 1 domain to both Jitsi Web loadbalancer.

You need to set ssl certificate there too for port 443 on setting of load balancer.

If you have more region just clone /shard1 to /shard2, update the region on 2 places:

  • For Jitsi web:

    /shard2/web-configmap.yaml

Find this and update

deploymentInfo: {
    shard: "shard1",
    region: "tor-1",
    userRegion: "tor-1"
}
  • For JVBs:

    /shard2/jvb-statefullset.yaml Update OCTO_REGION

Then redo from step 3.

Testing

Please refer to this testing using this configuration: https://community.jitsi.org/t/jitsi-jvb-2-performance-2020-testing/83672