1
0
Fork 0
No description
Find a file
2019-10-16 12:06:27 +02:00
containers removing trailing slash from inputs 2019-10-16 12:06:27 +02:00
docs added license 2019-10-04 14:31:57 +02:00
kubernetes removed unused ssh service 2019-10-14 18:26:47 +02:00
scripts/phabtalk added host parameter 2019-10-16 11:54:38 +02:00
.gitignore initial import of configuration files 2019-10-04 10:26:20 +02:00
k8s_config initial import of configuration files 2019-10-04 10:26:20 +02:00
LICENSE added license 2019-10-04 14:31:57 +02:00
local_setup.sh Add 2 more jenkins nodes. 2019-10-14 12:22:25 +02:00
README.md link to results server 2019-10-10 11:01:17 +02:00

Overview

This repository contains the configuration files for the merge guards for the LLVM project. It configures a cluster of build machines that are used to check all incoming commits to the LLVM project.

Merge guards

TODO(@christiankuehnel): describe objective of merge guards

Cluster overview

The cluster consists of these services:

Jenkins-Phabricator integration

The Jenkins-Phabricator is based on the instructions provided with the Phabricator-Jenkins Plugin.

On the Phabricator side these things were configured:

On the Jenkins side:

  • in the Jenkins configuration page as explained in the instrucitons
  • in the build job

There is no backup of the credentials. If you need to change it, generate a new one and update it in Jenkins and Phabricator.

Playbooks

deployment to a clean infrastructure

General remarks:

  • GCP does not route any traffic to your services unless the service is "healthy". It might take a few minutes after startup before the services is classified as healthy. Until then you will only see some generic error message.

These are the steps to set up the build server on a clean infrastructure:

  1. Configure the tools on your local machine:
    ./local_setup.sh
    
  2. Delete the old cluster, if it still exists:
    cd kubernetes/cluster
    ./cluster_delete.sh
    
  3. Create the cluster:
    cd kubernetes/cluster
    ./cluster_create.sh
    
  4. Create the disk storage, if it does not yet exist:
    cd kubernetes/cluster
    ./disk_create.sh
    
  5. SSH into the VM instance mounting the volume, find the mount point and then set
    # go to the mount point of the volume
    cd  /var/lib/kubelet/plugins/kubernetes.io/gce-pd/mounts/jenkins-home
    # change the permissions
    sudo chmod a+rwx
    
  6. Push the docker images to gcr.io:
    cd containers/debian-testing-clang8
    ./build_deploy.sh
    
    cd ../jenkins-master
    ./build_deploy.sh
    
  7. Deploy the stack:
    cd kubernetes
    ./deploy.sh
    
  8. Configure it

handling SSH keys

The Jenkins server SSHs into the agents to start the agent application. Thus the master needs SSH access to the agent. To set this up:

  1. Create an SSH key pair locally with ssh-keygen.
  2. Copy the contents of id_rsa to the credentials section of the Jenkins UI.
  3. Configure the agent in the Jenkins UI to use the new SSH keys you just uploaded.
  4. Copy the contents of id_rsa.pub to containers/<agent dir>/authorized keys.
  5. Rebuild and deploy the agents.

While this works, it does not fell like the perfect solution. I'm happy to get better ideas on this.

creating basic authentication for reverse proxy

  1. create auth file, based on ingress-nginx documentation
    cd kubernetes/reverse-proxy
    htpasswd -c auth <username>
    # enter password at prompt
    # add more users as required
    kubectl create secret generic proxy-auth --from-file=auth --namespace=jenkins
    

License

This project is licensed unter the "Apache 2.0 with LLVM Exception" license. See LICENSE for details.