When executing the command "mount" I can see this line:

/dev/sda1 on /home/jenkins type ext4 (rw,nosuid,nodev,noexec,relatime,commit=30,data=ordered)

It could be the origin of the issue because of the "noexec". How can I fix this simply since I'm not managing slave creation (due to Kubernetes plugin) please?

And even if I could fix that by overriding this mount or whatever... it seems strange this is not working from basic configuration of Kubernetes plugin, no?

The kubernetes plugin does not control the mount options of the temp folder. I think you'll need to work with your administrators to change the cluster configuration.

Another alternative is to use the JGit provider from the git plugin instead of command line git.

See the CloudBees article for more details.

markewaite I saw this article few hours ago but they are some problems:

• They are already with a 755 chmod. It doesn't work due to the mount having the parameter "noexec"
• And even if I remount partition be myself, my jenkins-slave are temporary due kubernetes plugin so I'm looking for a automatic way to do that. I guess I need to modify a setting in pod deployment YAML but I don't see at this moment where should I do that

If you have some ideas/indications, don't hesitate to give them

Thanks

sneko use JGit. The Jenkins git plugin includes support for command line git (default) and for JGit (needs to be enabled in the Jenkins global configuration by adding JGit as an option). Once you've enabled JGit globally in Jenkins, a pick list will appear in the job definition to allow you to choose the git implementation.

I'm gonna try tomorrow, it's 4am here. I will keep you updated and will mark as resolved if I succeed

Thanks for your help

Hahaha... that's was so simple to solve this issue by using JGit... Big thank you markewaite

By the way I will try to figure it out with docker image maintainer since /home/jenkins is the only volume I provide from Kubernetes and that by executing "mount" I can get this list:

/dev/sda1 on /home/jenkins type ext4 (rw,nosuid,nodev,noexec,relatime,commit=30,data=ordered)
/dev/sda1 on /home/jenkins/agent type ext4 (rw,nosuid,nodev,relatime,commit=30,data=ordered)
/dev/sda1 on /home/jenkins/.jenkins type ext4 (rw,nosuid,nodev,relatime,commit=30,data=ordered)

So I just looked into the slave Dockerfile (https://hub.docker.com/r/jenkins/slave/~/dockerfile/) to see why some subfolders don't have noexec options... It appears that they are created during the docker image build. I'm not totally confident that I'm understanding the whole thing but we could imagine that by creating /home/jenkins/workspace during image build we would avoid the "noexec" option.

After that I don't have enough knowledge about Jenkins processes to know if it's secure and if it makes sense...

Thanks again you were very helpful 

See in comments one of the solution provided by markewaite

I just want to update the solution found since it was not scoped to Git, but also to all scripted parts written in my Jenkinsfiles.

By following these instructions:
https://cloud.google.com/solutions/configuring-jenkins-kubernetes-engine
 
You will use volumes from host, that apply for using docker and to locate the workspace in the /tmp of the host.
 
By doing so, it solves all the problems related to the "noexec" mount option on slave partitions.
 
I hope it will help!