activiti-cloud-full-chart

Getting Started Guide

More information:

• all chart archives, located at: https://github.com/Activiti/activiti-cloud-helm-charts
• full chart, located at: https://github.com/Activiti/activiti-cloud-full-chart (this repo)
• a common chart as a base chart for all charts, located at: https://github.com/Activiti/activiti-cloud-common-chart
• charts for components, as sub folders located at: https://github.com/Activiti/activiti-cloud-application

Running on Docker Desktop

Install Docker Desktop and make sure the included single node Kubernetes cluster is started.

Install the latest version of Helm.

Add the magic host.docker.internal hostname to your hosts file:

sudo echo "127.0.0.1        host.docker.internal" > /etc/hosts

Install a recent version of ingress-nginx:

helm install --repo https://kubernetes.github.io/ingress-nginx ingress-nginx ingress-nginx

Update all dependencies:

helm dependency update charts/activiti-cloud-full-example

Create Activiti Keycloak Client Kubernetes secret in the activiti namespace:

kubectl create secret generic activiti-keycloak-client \
   --namespace activiti \
   --from-literal=clientId=activiti-keycloak \
   --from-literal=clientSecret=`uuidgen`

Create a values.yaml file with any customised values from the default values.yaml you want, as documented in the chart README.

In your local installation to start with, this would be:

global:
  gateway:
    host: host.docker.internal
  keycloak:
    host: host.docker.internal
    clientSecretName: activiti-keycloak-client
    useExistingClientSecret: true

Alternatively, you can create Activiti Keycloak Client Kubernetes secret with Helm with the following values:

global:
  gateway:
    host: host.docker.internal
  keycloak:
    host: host.docker.internal
    clientSecret: changeit

In a generic cluster install, you can just add --set global.gateway.domain=$YOUR_CLUSTER_DOMAIN to the helm command line, provided your DNS is configured with a wildcard entry *.$YOUR_CLUSTER_DOMAIN pointing to your cluster ingress.

Install or upgrade an existing installation:

helm upgrade --install \
  --atomic --create-namespace --namespace activiti \
  -f values.yaml \
  activiti charts/activiti-cloud-full-example

Uninstall:

helm uninstall --namespace activiti activiti

WARNING All the PVCs are not deleted by helm uninstall and that should be done manually unless you want to keep data for another install.

kubectl get pvc --namespace activiti
kubectl delete pvc --namespace activiti ...

or just delete the namespace fully:

kubectl delete ns activiti

As an alternative, generate a Kubernetes descriptor you can analyse or apply offline using kubectl apply -f output.yaml:

helm template --validate \
  --atomic --create-namespace --dependency-update --namespace activiti \
  -f values.yaml \
  activiti charts/activiti-cloud-full-example

Enabling message partitioning

In order to enable partitioning provide the following extra values (partitionCount defines how many partitions will be used and the Helm deployment will create that many replicaSets of query service and configure Rb service with the number of supported partitions in Query):

global:
  messaging:
    # global.messaging.partitioned -- enables partitioned messaging in combination with messaging.enabled=true && messaging.role=producer|consumer
    partitioned: true
    # global.messaging.partitionCount -- configures number of partitioned consumers
    partitionCount: 4

Use Kafka instead of Rabbit MQ

In order to switch the message broker to Kafka add the following extra values

global:
  messaging:
    broker: kafka
kafka:
  enabled: true
rabbitmq:
  enabled: false

Message Partitioning with Kafka

Kafka has different architecture from RabbitMQ. One Kafka topic can be served by a number of partitions independent from the consumer number (greater or equal).

Configuring the Kafka broker in the helm chart it is possible to specify partitionCount greater or equal to the replicaCount (the consumer number). Defining these two number independently allow the user to instantiate consumers only if it is needed, avoiding to waste resources.

global:
  messaging:
    partitioned: true
    # global.messaging.partitionCount -- set the Kafka partition number
    partitionCount: 4

activiti-cloud-query:
  # replicaCount -- set the Kafka consumer number
  replicaCount: 2

Enabling HorizontalPodAutoscaler (HPA)

Kubernetes supports horizontal scalability through Horizontal Pod Autoscaler (HPA) mechanism. In activiti-cloud-full-charts it is now possible to enable HPA for the runtime-bundle and activiti-cloud-query microservices.

Requirements

The HorizontalPodAutoscaler can fetch metrics from aggregated APIs that, for Kubernetes (metrics.k8s.io), are provided by an add-on named Metrics Server.

So, Metric Server needs to be installed and launched to use the HPA feature. Please refer to this page for its installation.

HPA Configuration

In the activiti-cloud-full-chart the HorizontalPodAutoscaler is disabled by default for backward compatibility. Please add the following configuration to your values.yaml to enable and use it:

runtime-bundle:
  hpa:
    enabled: true
    minReplicas: 1
    maxReplicas: 6
    cpu: 90
    memory: "2000Mi"
activiti-cloud-query:
  hpa:
    enabled: true
    minReplicas: 1
    maxReplicas: 4
    cpu: 90

This configuration (present in the hpa-values.yaml file in this repository) enable HPA for both runtime-bundle and activiti-cloud-query.

⚠️ WARNING: the provided values are just an example. Please adjust the values to your specific use case.

Configuration Properties

Name	Description	Default
enabled	enables the HPA feature	false
minReplicas	starting number of replicas to be spawned
maxReplicas	max number of replicas to be spawned
cpu	+1 replica over this average % CPU value
memory	+1 replica over this average memory value
scalingPolicesEnabled	enables the scaling policies	true

Scaling Polices

Scaling policies allow Kubernetes to stabilize the number of pods when there are swift fluctuations of the load. The scale-down policies are configured so that:

• only 1 pod can be dismissed every minute.
• only 15% of the number of pods can be dismissed every minute.
• the policy that scales down more pods will be triggered first.

The scale-up policies are the default Kubernetes ones.

These policies are always enabled until a scalingPolicesEnabled: false is specified in the configuration.

Activiti Cloud Query and HPA

Activiti Cloud supports both RabbitMQ and Kafka message broker. Activiti Cloud Query is a consumer of the message broker, so we need to be extra careful in the configuration of the automatic scalability in order to keep it working properly.

As a general rule, the automatic horizontal scalability for the query consumers should be enabled only when the Activiti Cloud has enabled partitioning.

Activiti Cloud Query and HPA with Kafka

In a partitioned installation, Kafka allows the consumers to connect to one or more partitions with the maximum ratio of 1:1 between partitions and consumers.

So when configuring HPA please don't specify the maxReplicas value greater than the partitionCount.

Activiti Cloud Query and HPA with RabbitMQ

When partitioning RabbitMQ the configuration will spawn one replica for every partition, so you should avoid activating the HorizontalPodAutoscaler in this case.

Skipping CI

If you want to skip running release pipeline stages, simply add [ci skip] to your commit message.

CI/CD

Running on Github Actions, requires the following environment variable to be set:

Name	Description
GITHUB_TOKEN	Github token for git service account
GITHUB_USER	Github user name for git service account
K8S_API_TOKEN	Kubernetes API token
K8S_API_URL	Kubernetes API url