Helm chart for DiracX

This helm chart is intended to be used in two ways:

• Development: The ./run_demo.sh script allows the infrastructure to be ran locally with docker+kind
• Production: TODO

Workflow

This chart can be used for 4 different installation type:

• demo/dev: we install everything and configure everything with pre-configured values (see below)
• prod: you already have a DIRAC installation with it's own DBs and everything, so you want to create a cluster, but bridge on existing external resources (like DBs)
• New: you start from absolutely nothing (no DIRAC), and you want to install all the dependencies
• New without dependencies: you start with nothing, but you want to use externally managed resources (like DB provided by your IT service)

Depending on the installation you perform, some tasks may be necessary or not. The bottom line is that to simplify the various cases, we want to be able to always run the initialization steps (like DB initialization, or CS initialization) but they should be adiabatic and non destructive.

To understand how the chart operates, see reference

What this chart contains

This chart contains the deployment for diracx and diracx-web, as well as dependencies:

• Mysql database
• OpenSearch database
• Dex and IAM as identity provider
• Minio as an object store for the SandboxStore
• OpenTelemetry (see details)

Intro to Kubernetes and Helm

Kubernetes (k8s) allows you to deploy containerized applications while letting you abstract the underlying infrastructure. The configuration of your application and how you run it is communicated to k8s via yaml files.

Helm allows to template these yaml files. Such a templated description of an application like diracx is called chart. Helm also allows to manage dependencies between charts. For example, the diracx application needs a database to run, so the diracx charts has a dependency on the mysql charts.

To understand this chart you will need to familiarize yourself with a few k8s concepts:

• A node is one of the machine (VM, physical) which sustains your k8s cluster.
• Your application runs inside a container which is part of a pod. A pod is the atomic unit with which kubernetes will work, and in most cases it corresponds to a single container. pods are deployed on nodes.
• A ReplicaSet represents how many pods of a given type you want to run. For example, you want to run 3 pods containing the diracx container for redundancy reasons.
• A Deployment is how you describe your workload to k8s. It ensures that any number of replicas of your pod are running (via ReplicaSEt). This chart contains a deployment for diracx (see here), and a deployment for diracx-web (see here). Kubernetes will always make sure that the deployment is satisfied.
• A Service is how you expose your Deployment. If I want to talk to my diracx application, it is a Service which will take care of redirecting me within the cluster to one of the pod. Most of the time, the Service is used for routing inside the cluster
• An Ingress exposes your Services outside of the cluster.

Running the demo locally

Running an instance locally is useful for demo or testing purposes. This entails running a local kubernetes cluster, installing this helm chart and all its dependencies, and configuring it accordingly. In order to make that easy, we wrote the run_demo.sh script, which covers all these needs. This is also used for the diracx ci

run_demo.sh

This is the script used to start a demo or the integration tests.

Usage: run_demo.sh [-h|--help] [--exit-when-done] [--offline] [--enable-coverage] [--no-mount-containerd] [--set-value key=value] [--] [source directories]

  -h|--help: Print this help message and exit
  --exit-when-done: Exit after the demo has been started (it will be left running in the background)
  --enable-coverage: Enable coverage reporting (used by diracx CI)
  --no-editable-python: Do not install Python source directories in editable mode
  --offline: Run in a mode which is suitable for fully offline use.
             WARNING: This may result in some weird behaviour, see the demo documentation for details.
             Implies: --mount-containerd
  --no-mount-containerd: Mount a directory on the host for the kind containerd storage.
                         This option avoids needing to pull container images every time the demo is started.
                         WARNING: There is no garbage collection so the directory will grow without bound.
  --enable-open-telemetry: lauches OpenTelemetry collection.
                           WARNING: experimental and resource hungry.
  --set-value: Set a value in the Helm values file. This can be used to override the default values.
               For example, to enable coverage reporting pass: --set-value developer.enableCoverage=true
  source directories: A list of directories containing Python packages to mount in the demo cluster.

Once ready, you will get some information on how to interact with your cluster

NOTES:

🚧 Chart has been installed in developer mode.

ℹ️  To interact with the cluster directly using kubectl/helm you'll need to set the
first following environment variables:

  export KUBECONFIG=/home/chaen/dirac/diracx-project/diracx-charts/.demo/kube.conf
  export HELM_DATA_HOME=/home/chaen/dirac/diracx-project/diracx-charts/.demo/helm_data
  export PATH=${PATH}:/home/chaen/dirac/diracx-project/diracx-charts/.demo

Then see the chart README for more information on how to use kubectl/helm.

ℹ️  To use the demo with the DiracX client set:

  export DIRACX_URL=https://172.18.0.1.nip.io:8000
  export DIRACX_CA_PATH=/home/chaen/dirac/diracx-project/diracx-charts/.demo/demo-ca.pem

ℹ️  To use the demo with a DiracX client without local installation:

  kubectl exec -it  deployment/diracx-demo-cli -- bash

ℹ️  To access the web application visit:

  https://172.18.0.1.nip.io:8000

ℹ️  To access the interactive API documentation visit:

  https://172.18.0.1.nip.io:8000/api/docs

ℹ️  When prompted to login the credentials are:

  Username: admin@example.com
  Password: password

ℹ️  Grafana dashboard for telemetry (if enabled)

  https://172.18.0.1.nip.io:32004

  Username: admin
  Password: run ``kubectl get secrets diracx-demo-grafana -o jsonpath="{.data.admin-password}" | base64 --decode ; echo``

🦄 Waiting for installation to finish...
pod/diracx-demo-7fd5c47cd6-mq8s7 condition met
🎉 🎉 🎉 Pods are ready! 🎉 🎉 🎉
🦄 Creating initial CS content ...
Writing back configuration to /cs_store/initialRepo/default.yml
Successfully added vo to git+file:///cs_store/initialRepo
Writing back configuration to /cs_store/initialRepo/default.yml
Successfully added user to git+file:///cs_store/initialRepo

ℹ️  Press Ctrl+C to clean up and exit

Interacting with the demo

Kubernetes basics

Assuming you have exported the environment variables printed by the demo script you can interact with the demo cluster using:

# List the running pods
kubectl get pods
# Get some more information about a pod
kubectl describe pod/<pod name>
# Show the logs of a running pod
kubectl logs <pod name>
# Show the logs of a running pod and keep following them
kubectl logs -f <pod name>
# Run a command in one of the pods
kubectl exec -it <pod name> -- /bin/bash /entrypoint.sh

Helm basics

When running the demo, an values.yaml file is created as .demo/values.yaml. This file can be used with helm to interact with the running demo to make changes to what is running in the cluster.

# Show what will be changed by running "helm upgrade"
# Note: helm diff has a ``bug`` that requires you to specify the existing password
# https://github.com/databus23/helm-diff/issues/460

export RABBITMQ_PASSWORD=$(kubectl get secret --namespace "default" rabbitmq-secret -o jsonpath="{.data.rabbitmq-password}" | base64 -d)
export MYSQL_ROOT_PASSWORD=$(kubectl get secret --namespace "default" mysql-secret -o jsonpath="{.data.mysql-root-password}" | base64 -d)
export MYSQL_PASSWORD=$(kubectl get secret --namespace "default" mysql-secret -o jsonpath="{.data.mysql-password}" | base64 -d)

helm diff upgrade diracx-demo  ./diracx --values .demo/values.yaml --set rabbitmq.auth.password=$RABBITMQ_PASSWORD  --set mysql.auth.rootPassword=$MYSQL_ROOT_PASSWORD --set mysql.auth.password=$MYSQL_PASSWORD

# Actually run "helm upgrade" to apply changes
helm upgrade diracx-demo ./diracx --values .demo/values.yaml

See here for more details on what you can do to alter the behavior of the local installation.

Deploying in production

TODO: Link to k3s

TODO: Explain how to download the values from helm

TODO: add info about diracx-web

Deploying a custom branch to DIRAC certification

Apply the following on top of the standard values.yaml file, replacing USERNAME and BRANCH_NAME with the appropriate values.

global:
  images:
    tag: "dev"
    # TODO: We should use the base images here but pythonModulesToInstall would need to be split
    services: ghcr.io/diracgrid/diracx/services
    client: ghcr.io/diracgrid/diracx/client

diracx:
  pythonModulesToInstall:
    - "git+https://github.com/USERNAME/diracx.git@BRANCH_NAME#egg=diracx_core&subdirectory=diracx-core"
    - "git+https://github.com/USERNAME/diracx.git@BRANCH_NAME#egg=diracx_db&subdirectory=diracx-db"
    - "git+https://github.com/USERNAME/diracx.git@BRANCH_NAME#egg=diracx_routers&subdirectory=diracx-routers"yaml

OpenTelemetry

:warning: Experimental: opentelemetry is an evolving product, and so is our implementation of it.

diracx aim at relying on OpenTelemetry for traces, monitoring and logging. When running in demo mode, this chart can spawn the necessary component for the telemetry to be reported:

• OpenTelemetry-collector to collect all the data
• Prometheus for the metrics
• Jaeger for traces
• ElasticSearch for logs (OpenSearch not yet supported)
• Grafana to display all that (accessible on port 32004 of the demo)

To enable it, run run_demo.sh with enable-open-telemetry

Note that this configuration is trivial and does not follow production recommandations (like using batch processing)

Requirements

Values

Autogenerated from chart metadata using helm-docs v1.11.2