Come for the products,
stay for the community

The Atlassian Community can help you and your team get more value out of Atlassian products and practices.

Atlassian Community about banner
Community Members
Community Events
Community Groups

Run two kubectl commands in a pipe, having output of the first one as an input for the second

For apply / replace configMap I need to run two kubectl commands in the following way:


kubectl create configmap sb-products-files --from-file=configmap_files/ \

   -o yaml --dry-run=client | kubectl replace --namespace=myNamespace -f -

Basically, I need to run one command, and it's output send as input to another command.

Is there any way to accomplish it in bitbucket pipeline?

2 answers

I've got similar issue. I need to execute `kubectl` to generate a manifest file on the fly. 

kubectl create configmap foo --from-file=foo.txt -o yaml --dry-run > foo.yaml

Using atlassian/kubectl-run would require me to configure and provide KUBE_CONFIG (because it's mandatory for that pipe) which I'd like to avoid. 

- pipe: atlassian/kubectl-run:3.1.2
    KUBE_CONFIG: '<string>'
    KUBECTL_COMMAND: '<string>'

So I ended up adding an extra step with custom image to generate output file which I pass to the next step through as an artifact.

- step: 
name: 'Create ConfigMap'
image: bitnami/kubectl
- foo.yaml
- kubectl create configmap foo --from-file=foo.txt -o yaml --dry-run > foo.yaml

For such cases, lifting requirement for mandatory KUBE_CONFIG in the official pipe would be practical (or to add `kubectl` to atlassian/default-image).

0 votes

@Kostya Kostyushko you could execute create configmap firstly and save it to file, and then run a pipe with RESOURCE_PATH equal to that file you saved output to

Regards, Galyna

Can you give examples please? I tried to make something, but it didn't work.

@Kostya Kostyushko I meant somethong like

- kubectl ... > output.yml
- pipe: atlassian/kubectl-run:2.0.0
RESOURCE_PATH: 'output.yaml'

The location of your resource file can vary depending on your requirements and project structure.

Regards, Galyna

bash: kubectl: command not found


I'm using

Images used:    build :

Ok. I found.
For those who need to run multiple kubectl commands
this will be a good solution:


      - step:
          # trigger: manual
          name: Deploy to Kubernetes
          image: atlassian/pipelines-kubectl
            # NOTE: $KUBECONFIG is secret stored as a base64 encoded string
            # Base64 decode our kubeconfig file into a temporary kubeconfig.yml file (this will be destroyed automatically after this step runs)
            - echo $KUBECONFIG | base64 -d > kubeconfig.yml
            # Tell our Kubernetes deployment to use the new Docker image tag
            - kubectl --kubeconfig=kubeconfig.yml --namespace=<namespace> set image deployment/<deployment-name> <deployment-name>=<docker-username>/<docker-image>:$BITBUCKET_COMMIT


by using  image: atlassian/pipelines-kubectl

Like guik likes this

@Kostya Kostyushko yeap, you can use also kubectl official version, whatever is suitable for you

In the future we may think about suppoorting output sharing from pipe execution, so that you can execute pipes sequentially passing output from previous execution

Suggest an answer

Log in or Sign up to answer

Atlassian Community Events