K
Q

Helm - Release multi container application and partial release

11/29/2021

We have below chart structure in helm which can deploy 'a microservice' to our k8s cluster

service-chart
|_templates
  |_deployment.yaml
  |_Ingress.yaml
  |_service.yaml
|_Chart.yaml
|_valueBase.yaml
|_valueForService1.yaml
|_valueForService2.yaml
..
..
|_valueForService{n}.yaml

valueBase.yaml contains default values for a service (for eg. limits, replicas etc)

global:
 namespace: teamname
 environment: staging
limits:
 memory: "500Mi"
 cpu: "300m"
deployments:
 replicas: 1
probes:
 path: "/"

and valueForService1.yaml file contains values overridden for a specific service

app: 
 name: "service1"
image:
 name: "service1"
 version: "2021.11.xxx"
port: 50001
deployments: 
 replicas: 3

All of above services follow exactly same structure and create similar resources i.e. a service, a pod and an ingress. we deploy an individual service as

helm upgrade --install -f valueBase.yaml -f valueForService1.yaml service1 .

But problem is we have 50 of these services. and we would want to install all of them together instead of running 50 commands back to back. Also would like to have some release co-ordination in between releasing multiple services. for eg. release service1 before releasing service2.

I know we can do this in one command with 50 -f parameters in it but that's not the kind of solution I am here for.

How can we package them properly so we can

  • release all services at once when we want to.
  • release an individual service
  • release them as group of service for eg. group1 consist of service1, service2 and service3
  • release orchestration for eg. release service1 then service2 then service3

All suggestions welcomed. Feel free to ask for more details.

Note: We tried using sub-chart but it doesn't appear to be solved by using sub-charts. We just have lots of services with similar structure. but I might have used sub charts completely wrong.

-- ThePravinDeshmukh
helm3
kubernetes
kubernetes-helm
yaml

Similar Questions

When using Istio on EKS, how can I restrict the IPs in the security group rules that are created?
Deploy WAR in Tomcat on Kubernetes
Spring Boot Actuator + Spring Boot Admin - Is there a way to define a custom management url?
In a Kubernetes cluster, is there a way to migrate etcd from external to internal?
Kafka partition reassignment with horizontal pod autoscaling

1 Answer

11/29/2021

What you're looking for is helmfile

  1. Store your base chart and services values in the same folder.

  2. Create helmfile.yml.

    releases:
  - name: service1
    labels:
      app: service1
      group: group1
    chart: charts/baseChart
    values:
      - valueBase.yaml
      - valueForService1.yaml
      - fullnameOverride: service1
  - name: service2
    labels:
      app: service2
      group: group1
    chart: charts/baseChart
    values:
      - valueBase.yaml
      - valueForService2.yaml
      - fullnameOverride: service2
    needs:
      - service1

    Each release in the releases list is a helm release definition. You can control installation order with the needs keyword.

  3. Now your folder looks something like this

    .
├── charts/
│   └── baseChart/
├── helmfile.yml
├── valueBase.yml
├── valueForService1.yml
└── valueForService2.yml

    and you're ready to run helmfile commands

  • helmfile -n <namespace> apply will release all services at once

  • helmfile -n <namespace> -l app=service1 apply will release an individual service with the app: service label

  • helmfile -n <namespace> -l group=group1 apply will release all services with the group: group1 label

-- zzamzam
Source: StackOverflow