Running in a jenkins slave kubernetes pod fails

[tculp]

Using this image in a jenkins slave pod fails.

Here is an example jenkinsfile (using the Kuberneted plugin):

podTemplate(
  containers: [
    containerTemplate(
      image: "curlimages/curl:7.70.0",
      name: 'curl',
      command: 'cat',
      ttyEnabled: true
    )
  ]
)
{
node(POD_LABEL){

  stage('Curl google'){
    container('curl') {
      sh "curl google.com"
    }
  }
}
}

Which fails with:

[Pipeline] stage
[Pipeline] { (Curl google)
[Pipeline] container
[Pipeline] {
[Pipeline] sh
process apparently never started in /home/jenkins/agent/workspace/network-debugger_master@tmp/durable-6e3552c0
(running Jenkins temporarily with -Dorg.jenkinsci.plugins.durabletask.BourneShellScript.LAUNCH_DIAGNOSTICS=true might make the problem clearer)
[Pipeline] }

Meanwhile, the following Jenkinsfile works as expected:

podTemplate(
  containers: [
    containerTemplate(
      image: "centos:8",
      name: 'curl',
      command: 'cat',
      ttyEnabled: true
    )
  ]
)
{
node(POD_LABEL){

  stage('Curl google'){
    container('curl') {
      sh "curl google.com"
    }
  }
}
}

[xquery]

we would need more logs to debug what is going on here eg. from what you have shown it looks like curl is never even invoked.

a random guess - you might try invoking the curlimages/curl entry script.

[xquery]

without any additional info (and not knowing the specifics of your jenkins & kubernetes environment) it is hard to see how we can help - I am going to close this issue but will try a few random guesses (feel free to comment and I will reopen if you think it would help).

running

> docker run --rm -it curlimages/curl which curl

shows curl exists at /usr/bin/curl so should be on the $PATH

Note that curl in the docker container runs under the curl_user (thats an explicit design decision) ... it is unclear in jenkins/kubernetes if a user is being specified.

Maybe the centos version has a ./curlrc file amending curl behaviour eg. with that particular uri you would need to supply -L to ensure redirects are being followed eg.

> docker run --rm -it curlimages/curl curl -L google.com

[cmamigonian]

I am seeing this too, and maybe I can explain a little more. The pod uses the curl image and runs cat as a noop just to start the container, then within the Jenkins stage step--which is running inside the curl container in the pod--we run the curl command in a shell. It appears as if running cat in the container is just hanging the shell so we never actually get to executing the curl command. I also notice this if I pull the image locally and run cat: it just hangs.

[tculp]

I do see that it hangs forever locally with a cat call. However, using any other image that I've encountered, such as ubuntu or alpine, does the same, and successfully works in a Jenkins + Kube setup. I'm not sure what the difference is with this image.

[wayne0811]

Ran into this issue as well. Eventually I found that this is documented in the Kubernetes plugin: https://github.com/jenkinsci/kubernetes-plugin#pipeline-sh-step-hangs-when-multiple-containers-are-used

Any of the following fixes worked for me:

• Add runAsUser: '1000' to the podTemplate
• Add runAsUser: '1000' to the curl containerTemplate
• Add runAsUser: '0' to the curl containerTemplate

[alpeshjikadra]

I was facing the same issue from today morning,

As a solution I have recreated docker image which was getting used in container with USER 1000 option. and it started working

i.e. Dockerfile

`FROM curlimages/curl:7.78.0

USER 1000`