Docker runner ephemeral disk full

Drone Support
1

Im using helm chart stable on Amazon EKS
Drone is doing a good job with all my builds. Only with one project that the image is 10GB, it fails to build. I see kubernetes setting the agent to Evicted with error: ephemeral disk space full.

I tried to follows a few examples from other support tickets but with no luck.

I have this simple pipeline below:

kind: pipeline
type: docker
name: default

steps:
  - name: publish
    image: plugins/ecr
    environment:
      PLUGIN_STORAGE_DRIVER: overlay2
      PLUGIN_CACHE_FROM: test.xxxxx.xxx.eu-west-1.amazonaws.com/test:latest
    settings:
      repo: test
      region: eu-west-1
      tags:
        - ${DRONE_COMMIT}
        - latest
      access_key:
        from_secret: aws_access_key_id
      secret_key:
        from_secret: aws_secret_access_key
      registry: test.xxxxx.xxx.eu-west-1.amazonaws.com

  - name: deploy
    image: quay.io/hectorqin/drone-kubectl
    settings:
      kubernetes_template: k8s/api-deployment.yaml
      kubernetes_namespace: default
      kubernetes_server: https://test.xxxxxx.xxxx.sk1.eu-west-1.eks.amazonaws.com
      kubernetes_cert:
        from_secret: KUBERNETES_CERT
      kubernetes_token:
        from_secret: KUBERNETES_TOKEN
    environment:
      DRONE_COMMIT: ${DRONE_COMMIT}

trigger:
  branch:
    - development

What is the best way of having more space on my DinD or maybe should I use another runner? I havent use Kubernetes runner as it is deprecated.

I can give more info if needed, thanks

2

speaking only for myself, I have no clue how space is allocated in the dind image. You may consider posting to Docker community forum since they generally have better knowledge of the docker:dind image (on which the plugin is based) or alternatively forking the plugin and modifying for your particular needs.

I havent use Kubernetes runner as it is deprecated.

the experimental kubernetes runner was in fact deprecated. However a second attempt at a kubernetes runner was published for preview. See Overview | Drone

3

I will do a test with Kubernetes Runner then.
I still dont understand the part of mounting volumes on the pipeline step to make it available on the agent for the image building