-
Bug
-
Resolution: Duplicate
-
Major
-
None
-
CloudBees Docker Pipeline 1.9
Jenkins is running inside a docker container
Docker Server is accessed via tcp (TLS)
Jenkins home directory is mounted via volume
Using image.inside() inside docker.withServer() does fails with the following error:
[Pipeline] withDockerContainer $ docker run -t -d -u 1000:1000 -w "/var/jenkins_home/workspace/Utility Jobs/copy-schema" -v "/var/jenkins_home/workspace/Utility Jobs/copy-schema:/var/jenkins_home/workspace/Utility Jobs/copy-schema:rw" -v "/var/jenkins_home/workspace/Utility Jobs/copy-schema@tmp:/var/jenkins_home/workspace/Utility Jobs/copy-schema@tmp:rw" -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** --entrypoint cat dockerimage [Pipeline] { [Pipeline] sh [copy-schema] Running shell script sh: /var/jenkins_home/workspace/Utility Jobs/copy-schema@tmp/durable-f8d02fd0/pid: No such file or directory sh: /var/jenkins_home/workspace/Utility Jobs/copy-schema@tmp/durable-f8d02fd0/jenkins-log.txt: No such file or directory sh: /var/jenkins_home/workspace/Utility Jobs/copy-schema@tmp/durable-f8d02fd0/jenkins-result.txt: No such file or directory [Pipeline] } $ docker stop --time=1 d1f1d8173fcc89762825aeee7019aa24e88e4f9c87c08036083118cd4c708f26 $ docker rm -f d1f1d8173fcc89762825aeee7019aa24e88e4f9c87c08036083118cd4c708f26 [Pipeline] // withDockerContainer
I've debugged my Jenkins instance and it turns out that the docker inspect calls inside org.jenkinsci.plugins.docker.workflow.WithContainerStep.Execution.start() fail silently because the docker client can't connect to the docker host. The reason seems to be that the call to org.jenkinsci.plugins.docker.workflow.client.DockerClient.getVolumes(EnvVars, String) uses envHost which does not contain the environment variables set by withServer(). If I pass in all environment variables ( env) the call succeeds and the inside() step succeeds as well (the volume from the master is reused using --volumes-from).
I'm not sure about the other implications of passing in env instead of envHost.
- duplicates
-
JENKINS-43222 DOCKER_HOST set on slave ignored by docker-pipeline
-
- Resolved
-
- is duplicated by
-
-
- Resolved
-
- relates to
-
JENKINS-36348 Demo improvements
-
- Open
-
- links to
[JENKINS-39243] image.inside() does not respect docker.withServer() settings
Craig Rodrigues
added a comment - cfraenkel can you post an example of how you passed the value of DOCKER_HOST down so that image.inside() works with the settings specified in docker.withServer() ?
I'm trying to do the same thing you are doing, where my Docker server is on a separate host from where my Jenkins job is running.
Craig Rodrigues
added a comment - cfraenkel can you post an example of how you passed the value of DOCKER_HOST down so that image.inside() works with the settings specified in docker.withServer() ?
I'm trying to do the same thing you are doing, where my Docker server is on a separate host from where my Jenkins job is running. rodrigc: I've recompiled the plugin using the attached patch (JENKINS-39243_env.patch
). As stated above, this may expose (far) too many environment variables though.
Christian Fraenkel
added a comment - - edited rodrigc : I've recompiled the plugin using the attached patch ( JENKINS-39243_env.patch ). As stated above, this may expose (far) too many environment variables though. I'm experiencing a very similar problem and using the changes in the attached patch don't seem to fix my issue. I used the latest commit to master when I tested against docker-workflow-plugin (eb4193c33d0a1a3c1f09b4197b5052729f97dcbb).
My setup is very simple...
- 1 docker host
- 1 jenkins master
On the Jenkins master, I have an env var set to:
DOCKER_HOST=tcp://10.x.x.x:4243
On the docker host, I have it listening over the same tcp port:
OPTIONS="-H tcp://0.0.0.0:4243"
No firewall rules are in the way and I can reach the other host with docker -H tcp://10.x.x.x:4243 ...
When I trigger a build, docker is invoked and will pull down a recent image on the external docker host, but once the docker.image().inside() function is called, I get an identical error to the OP.
When I install docker directly onto the Jenkins master, add jenkins to the docker group and unset the DOCKER_HOST env var, it works as expected but I obviously don't want the docker daemon running on the master.
Jenkins Core: 2.32.3
Pipeline: 2.5
Durable Task Plugin: 1.13
Docker Pipeline: 1.11-SNAPSHOT
Sounds plausible. Would need to first strengthen the demo to actually run (and rely on) withServer, which perhaps means ceasing to directly bind-mount docker.sock, and switching to creating some kind of TCP proxy (ideally behind TLS) that would in turn connect to the host daemon.