jenkinsci/docker

Derived jenkins docker image fails during init phase on helm install with `exec /usr/bin/sh: exec format error`

hectoralicea opened this issue · 3 comments

hectoralicea commented

Jenkins and plugins versions report

Environment 
Paste the output here

What Operating System are you using (both controller, and any agents involved in the problem)?

Running in Amazon Linux in an EKS cluster
based (FROM) derived docker image from: docker.io/jenkins/jenkins:2.375.2-rhel-ubi8-jdk11

Reproduction steps

  1. ) Build the a derived docker image using the following docker source, named: XXXXXXXXXXX.dkr.ecr.us-gov-west-1.amazonaws.com/das/gw/jenkins/jenkins
FROM docker.io/jenkins/jenkins:2.375.2-rhel-ubi8-jdk11
ARG user=jenkins

USER root
# make custom changes to jenkins container for this customer

RUN dnf update -y && \
    dnf -y -q install ca-certificates jq bind-utils nmap && \
    rm -rf /var/cache /var/log/dnf* /var/log/yum.* && \
    dnf clean all -y

COPY --chown=${user}:${user} daas-sandbox-root-ca.crt /etc/pki/ca-trust/source/anchors/daas-sandbox-root-ca.crt

RUN chmod 644 /etc/pki/ca-trust/source/anchors/daas-sandbox-root-ca.crt && \
    cat /etc/pki/ca-trust/source/anchors/daas-sandbox-root-ca.crt && \
    echo run update-ca-trust && \
    update-ca-trust && \
    echo ran update-ca-trust
    
USER ${user}

# as specified in the jenkins Docker file, 
# from a derived Dockerfile, can use `RUN install-plugins.sh active.txt` to setup $REF/plugins from a support bundle
RUN install-plugins.sh active.txt

ENTRYPOINT ["/sbin/tini", "--", "/usr/local/bin/jenkins.sh"]

2.) create a values.yaml file to feed to the helm deploy which has the following settings

controller:
  image: "XXXXXXXXXXX.dkr.ecr.us-gov-west-1.amazonaws.com/das/gw/jenkins/jenkins"
  tag: "2.375.2-rhel-ubi8-jdk11"

run a helm install

Jenkins should startup without issue.

Expected Results

Jenkins should startup without issue. Init container should install all the plugins, then the jenkins pod should boot strap and start the jenkins app.

Actual Results

It fails at container init. The following error shows in the log

hector$ kc logs -n cicd jenkins-0 -c init
exec /usr/bin/sh: exec format error

hector$ kc get pods -n cicd
NAME        READY   STATUS                  RESTARTS         AGE
jenkins-0   0/2     Init:CrashLoopBackOff   46 (2m47s ago)   3h34m

The events from this command is
kc describe pod -n cicd jenkins-0

Events:
  Type     Reason   Age                      From     Message
  ----     ------   ----                     ----     -------
  Normal   Pulling  34m (x41 over 3h34m)     kubelet  Pulling image "XXXXXXXXXXX.dkr.ecr.us-gov-west-1.amazonaws.com/das/gw/jenkins/jenkins:2.375.2-rhel-ubi8-jdk11"
  Warning  BackOff  4m11s (x969 over 3h34m)  kubelet  Back-off restarting failed container

Anything else?

What am I doing wrong in my derived Dockerfile?

dduportal commented

Hi @hectoralicea , what is the kind of CPU used for the nodes in your Kubernetes cluster?

The errors sounds like an Intel (linux/amd64) Docker image executed on an ARM (linux/arm64) machine (but I could be wrong).

hectoralicea commented

yea, I just realized this. I'm building this on a M1 MacBook, and but running on an EKS cluster inside an Amazon Linux which I presume is x86-64. Need to find the docker build parameter which instructs to build an linux/x86-64 architecture.

dduportal commented

yea, I just realized this. I'm building this on a M1 MacBook, and but running on an EKS cluster inside an Amazon Linux which I presume is x86-64. Need to find the docker build parameter which instructs to build an linux/x86-64 architecture.

docker build --platform=linux/amd64 <...> is your friend as per https://docs.docker.com/engine/reference/commandline/build/ 🤗

Be warned that it might be slower on your machine (because QEMU or Rosetta will take care of emulating the amd64 platform): be patient 😅

I'm closing the issue as it's not a problem with the Jenkins Docker image.