SciLifeLab Serve

SciLifeLab Serve (https://serve.scilifelab.se) is a platform offering machine learning model serving, app hosting (Shiny, Streamlit, Dash, etc.), web-based integrated development environments, and other tools to life science researchers affiliated with a Swedish research institute. It is developed and operated by the SciLifeLab Data Centre, part of SciLifeLab. See this page for information about funders and mandate.

This repository contains Helm charts for SciLifeLab Serve.

Reporting bugs and requesting features

If you are using SciLifeLab Serve and notice a bug or if there is a feature you would like to be added feel free to create an issue with a bug report or feature request.

How to deploy

Prerequisites

A Kubernetes cluster version 1.28.6
Helm 3
A storage class for dynamic provisioning of persistent volumes

If you are going to run this on a remote cluster, then you probably don't need to think about this as these things will be provided by your cloud provider.

But in case of a local deployment, navigate to the next section.

Setup for local deployment

If you are going to run this locally, you need to have a Kubernetes cluster running on your machine. You can use Rancher Desktop for this purpose.

Follow their instruction to install Rancher Desktop, and then start it.

Recommended settings for Rancher Desktop:

Preferences > Kubernetes select kubernetes version 1.28.6.
Preferences > Container Engine select containerd as the container engine.
Preferences > Virtual Machine > Emulation select QEMU
- If you are running on an M3 Mac select VZ
Preferences > Virtual Machine > Hardware select 4 CPUs and 16 GB of memory.

Serve image

By default, the image is pulled from the public registry. This image is the one we are using in production. So you don't need to build the image yourself if you want to just try it out locally.

But if you want to develop, you need to build the image yourself.

Building image for Rancher Desktop

Rancher Desktop brings a number of tools when you install it. One of them is nerdctl which is a drop-in replacement for docker and docker-compose.

Rancher Desktop also brings a local registry that you can use to push images to. And this registry can be accessed from your Kubernetes cluster and used as if you were using docker.

See Serve repository for up-to-date instructions on how to build the image for local development.

But this setup expects that you have an image tagged mystudio built using nerdctl and pushed to the k8s.io namespace.

Deploying

Using the following you'll make sure that your Rancher Desktop installation is working as expected using the default settings. These instructions are almost the same as the ones you would use for a remote cluster except for the storage class. If it doesn't work you should debug your installation and contact team members for help.

Outcomes of this section

You'll prepare your environment for the proper local deployment of Serve;
Running instance of Serve on your local machine available on http://studio.127.0.0.1.nip.io/.

First, clone this repository

$ git clone https://github.com/ScilifelabDataCentre/serve-charts.git

Then navigate to the serve-charts/serve folder

$ cd serve-charts/serve

Now you need to create an override file for the values.yaml file.

Create a file called values-local.yaml and add the following content:

# https://helm.sh/docs/chart_template_guide/yaml_techniques/#yaml-anchors
# for local development
storageClass: &storage_class local-path
#storage access mode
access_mode: &access_mode ReadWriteOnce
accessmode: *access_mode

global:
  studio:
    superuserPassword: "Test@12345"
    superuserEmail: "[email protected]"
    storageClass: *storage_class
  postgresql:
    storageClass: *storage_class

studio:
  # Only locally on a debug environment
  debug: true
  storage:
    storageClass: *storage_class
  media:
    storage:
      storageClass: *storage_class
      accessModes: *access_mode

postgresql:
  primary:
    persistence:
      storageClass: *storage_class
      accessModes:
        - *access_mode

This is necessary because the default values are set for a production environment. Specifically, the storage class has to change because the default storage class is not available in a Rancher Desktop environment.

$ helm dependency update
# The following command will install the chart with the values from values.yaml and values-local.yaml
# values-local.yaml will override the values from values.yaml
$ helm install serve . -f values.yaml -f values-local.yaml

As a result you should have a running instance of Serve on your local machine available on http://studio.127.0.0.1.nip.io/.

Swapping default docker image with the one built locally

TJ;DR Just commands

$ git clone https://github.com/ScilifelabDataCentre/serve-charts.git
$ cd serve-charts/serve
$ cat <<EOF > values-local.yaml
environment: "local"
# Path will be mounted using rancher desktop to the /app path in the container
source_code_path: "/Users/nikch187/Projects/sll/serve"
# https://helm.sh/docs/chart_template_guide/yaml_techniques/#yaml-anchors
# for local development
storageClass: &storage_class local-path
#storage access mode
access_mode: &access_mode ReadWriteOnce
accessmode: *access_mode

global:
studio:
  superuserPassword: "Test@12345"
  superuserEmail: "[email protected]"
  storageClass: *storage_class
postgresql:
  storageClass: *storage_class

studio:
# Only locally on a debug environment
debug: true
storage:
  storageClass: *storage_class
media:
  storage:
    storageClass: *storage_class
    accessModes: *access_mode

# We use pull policy Never because see the following link:
# https://github.com/rancher-sandbox/rancher-desktop/issues/952#issuecomment-993135128
static:
  image: mystudio
  pullPolicy: Never

image:
  repository: mystudio
  pullPolicy: Never

securityContext:
  # Disables security context for local development
  # Essentially allow the container to run as root
  enabled: false

readinessProbe:
  enabled: false

livenessProbe:
  enabled: false

postgresql:
primary:
  persistence:
    storageClass: *storage_class
    accessModes:
      - *access_mode 
EOF
$ helm upgrade serve . -f values.yaml -f values-local.yaml

Outcomes of this section:

Instead of a Django server, you'll have an ssh server running for the PyCharm setup
You'll have a host machine's folder with the Serve code mounted to the container;

Now that everything is running, you can swap the default image with the one you built locally.

See the Serve image section for instructions on how to build the image.

Go back to the values-local.yaml file update it with the following content:

environment: "local"

# Path will be mounted using rancher desktop to the /app path in the container
source_code_path: "/absolute/path/to/your/serve"
# https://helm.sh/docs/chart_template_guide/yaml_techniques/#yaml-anchors
# ...
studio:
  # Append the following to the end of the studio section
  
  # We use pull policy Never because see the following link:
  # https://github.com/rancher-sandbox/rancher-desktop/issues/952#issuecomment-993135128
  static:
    image: mystudio
    pullPolicy: Never

  image:
    repository: mystudio
    pullPolicy: Never

  securityContext:
    # Disables security context for local development
    # Essentially allow the container to run as root
    enabled: false

  readinessProbe:
    enabled: false

  livenessProbe:
    enabled: false

Full content of the values-local.yaml file

  environment: "local"
  # Path will be mounted using rancher desktop to the /app path in the container
  source_code_path: "/Users/nikch187/Projects/sll/serve"
  # https://helm.sh/docs/chart_template_guide/yaml_techniques/#yaml-anchors
  # for local development
  storageClass: &storage_class local-path
  #storage access mode
  access_mode: &access_mode ReadWriteOnce
  accessmode: *access_mode

  global:
    studio:
      superuserPassword: "Test@12345"
      superuserEmail: "[email protected]"
      storageClass: *storage_class
    postgresql:
      storageClass: *storage_class

  studio:
    # Only locally on a debug environment
    debug: true
    storage:
      storageClass: *storage_class
    media:
      storage:
        storageClass: *storage_class
        accessModes: *access_mode

    # We use pull policy Never because see the following link:
    # https://github.com/rancher-sandbox/rancher-desktop/issues/952#issuecomment-993135128
    static:
      image: mystudio
      pullPolicy: Never

    image:
      repository: mystudio
      pullPolicy: Never

    securityContext:
      # Disables security context for local development
      # Essentially allow the container to run as root
      enabled: false

    readinessProbe:
      enabled: false

    livenessProbe:
      enabled: false

  postgresql:
    primary:
      persistence:
        storageClass: *storage_class
        accessModes:
          - *access_mode

After doing this run the following command to upgrade the deployment:

helm upgrade serve . -f values.yaml -f values-local.yaml

Now you can proceed to set up PyCharm

If you don't want to set up PyCharm, you can just run Django from the container.

$ kubectl get po
# Get the name of the studio pod
$ kubectl exec -it <studio-pod-name> -- /bin/bash
# Now you are inside the container 
$ sh scripts/run_web.sh

Please note, that the folder you are in, /app, is the folder where the code is mounted.

It means that you can make changes to the code on your host machine and see the changes in the container.

Deploy an SSL certificate

For production you need a domain name with a wildcard SSL certificate. If your domain is your-domain.com, you will need a certificate for *.your-domain.com and *.studio.your-domain.com. Assuming that your certificate is fullchain.pem and your private key privkey.pem, you can create a secret prod-ingress containing the certificate with the command:

kubectl create secret tls prod-ingress --cert fullchain.pem --key privkey.pem

This secret should be in the same namespace as studio deployment.

Enabling network policies

If networkPolicy.enable = true, you have to make sure the correct kubernetes endpoint IP is provided in networkPolicy.kubernetes.cidr, and the correct port networkPolicy.kubernetes.port. This is to enable access of some services to the kubernetes API server through a created Service Account. To get your cluster's kubernetes endpoint run:

kubectl get endpoints kubernetes

To allow for within-cluster DNS, the kube-system namespace need the label:

kubectl label namespace kube-system name=kube-system

Further, for ingress resources you need to set networkPolicy.ingress_controller_namespace. If value can vary depending on your cluster configuration, but for NGINX ingress controller it's usually "ingress-nginx".

Example deployment

global:
  studio:
    superuserPassword: adminstudio # Django superuser password, username is admin
  postgresql:
      auth:
        username: studio
        password: studiopostgrespass
        postgresPassword: postgres
        database: studio
      storageClass: local-path 


networkPolicy:
  enable: true
  kubernetes:
    cidr: 127.0.0.1/32 # To get kubernetes api server endpoints run: $ kubectl get endpoints kubernetes
    port: 6443
  internal_cidr: # in-cluster IpBlock cidr, used in allow-internet-[egress|ingress] policy, e.g:
    - 10.0.0.0/8
    - 192.168.0.0/16
    - 172.0.0.0/20

studio:
  debug: false
  inactive_users: false #Users that sign-up can be inactive by default if desired
  csrf_trusted_origins: "https://studio.127.0.0.1.nip.io:8082" #extra trusted origin for django server, for example if you port-forward to port 8082
  image: # using a local image registry with hostname k3d-registry
    repository: k3d-registry:35187/serve:develop #This image can be built from Dockerfile (https://github.com/scaleoutsystems/serve)
    pullPolicy: Always # used to ensure that each time we redeploy always pull the latest image
  static:
    image: k3d-registry:35187/serve-nginx:develop #This image can be built from Dockerfile.nginx (https://github.com/scaleoutsystems/serve)
  media:
    storage:
      accessModes: ReadWriteOnce

accessmode: ReadWriteOnce

# Postgres deploy with a single-pod database:
postgresql:
  primary:
    persistence:
      size: "2Gi"
      accessModes:
        - ReadWriteOnce
      storageClass: local-path

rabbit:
  password: rabbitmqpass

redis:
  master:
    persistence:
      enabled: false
  replica:
    persistence:
      enabled: false

celeryFlower:
  enabled: false

reloader:
  enabled: true

Name		Name	Last commit message	Last commit date
Latest commit History 664 Commits
.github/workflows		.github/workflows
apps		apps
k8s-error-pages		k8s-error-pages
serve		serve
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

SciLifeLab Serve

Reporting bugs and requesting features

How to deploy

Prerequisites

Setup for local deployment

Serve image

Deploying

Swapping default docker image with the one built locally

Deploy an SSL certificate

Enabling network policies

Example deployment

About

Releases 47

Packages

Contributors 6

Languages

License

ScilifelabDataCentre/serve-charts

Folders and files

Latest commit

History

Repository files navigation

SciLifeLab Serve

Reporting bugs and requesting features

How to deploy

Prerequisites

Setup for local deployment

Serve image

Deploying

Swapping default docker image with the one built locally

Deploy an SSL certificate

Enabling network policies

Example deployment

About

Resources

License

Stars

Watchers

Forks

Releases 47

Packages 0

Contributors 6

Languages

Packages