Skip to content

Migration

Action

If you are currently using apps.ci.centos.org (OCP 3.6) or have a job running in ci.centos.org, please contact us so that we can create a namespace for you in the new OCP 4.4 cluster for you to start migrating your jobs.

If your own a namespace on apps.ci.centos.org

There shouldn't be a lot of changes for you as you should already have a familiar workflow working with Openshift, if you have configured your jobs at least once. While opening a ticket, or reaching out to us please do mention if you need ability to checkout duffy nodes (if yes, do you already have a duffy API key?).

See the following which explains how to open a ticket (Point 3)

If your project is on ci.centos.org

As you are using central jenkins deployment that's not private to you, your project might see some changes. You will have your own jenkins deployment to run all your jobs instead of having a shared jenkins instance where you don't have admin access. With the extra privileges in jenkins, you can add plugins, credentials as you need and a bunch of other abilities. This also means there is no direct maintenance required from CentOS CI Infra admins as openshift takes care of things dying or crashing and anything in the jobs config, you can take care of them yourself (We will help you get started and are always in reach if you notice any hiccups that is not auto healed). You can have the same workflow or use this opportunity to change your freestyle jobs to pipelines but that's totally on you. See the following which explains how to open a ticket (Point 3)

Open a project ticket

Create a new ticket in pagure.io/centos-infra/issues with type ci-migration template or click here.

We will have your accounts created in 48 hours (if no doubts/followup needed).

Note: This is only applicable for already existing projects in other places. New projects will have to go through a different process/evaluation and we may need more time (this is also applicable for projects requesting extra privileges than default admin level access to namespace).

Setting up your jobs in cico-workspace jenkins

Configuring your jobs should be similar to your older configuration. You can either have your jobs written in trigger jobs -> groovy dialogue box, or source it from your repo (we recommend the later for easy management of jenkinsfile). One thing to point, in groovy, the node parameter type defines where the target job should be executed, the value must match either a label or a node name - otherwise the job will just stay in the queue. We have a custom label called cico-workspace that has python-cicoclient installed in it so that you can request duffy nodes. This workspace will also have your duffy-api-key exported so that you can directly request nodes from duffy. Here is an example of jenkins file

node('cico-workspace') {
    stage('get cico node') {
        node = sh(script: "cico --debug node get -f value -c hostname -c comment", returnStdout: true).trim().tokenize(' ')
        env.node_hostname = "${node[0]}.ci.centos.org"
        env.node_ssid = "${node[1]}"
    }

    stage('tests-example') {
         println("Put your tests here")
         println("example running a test on ${node_hostname} ${node_ssid}")
    }

    stage('builds-example') {
        println("Put your builds here")
         println("example running a build on ${node_hostname} ${node_ssid}")
    }

    stage('return cico node') {
        sh 'cico node done ${node_ssid} > commandResult'
    }
}
Configuring the Kubernetes concurrency limit (number of cloud executors)

As the standard (static) Jenkins executors have been replaced by cloud-based ones, the limit is now configured in a different place. The configuration now resides under:

Manage Jenkins -> Manage Nodes and Clouds -> Configure Clouds -> Kubernetes -> Kubernetes Cloud details... -> Concurrency Limit

By default, the limit is set to 100, i.e. 100 parallel jobs at once. This default is in most scenarios way too high and can cause a quick Duffy pool depletion (if the spawned pods use the Duffy nodes). So, to be a good netizen, it is recommended to set this number to some sensible value (like 15 or 20, etc. - depends on your project), just to take in consideration the other users of the Duffy nodes & Kubernetes cluster.

Configuring SMTP server for email notifications

If you used job email notifications in the old (legacy) instance, you will probably find out that these don't work out-of-the box in the OCP instances. Fortunately, the solution is simple and consists of two steps:

  1. In Manage Jenkins -> Configure System in the Jenkins Location section fill in a valid System Admin e-mail address. Usually you want to use something like builder@\<your-Jenkins-instance-address> (which is also mentioned in the Jenkins URL field above). So, for example, the whole email address can be builder@my-fabulous-app.apps.ocp.ci.centos.org (the builder part is optional, you can use whatever you want).

  2. Again, in Manage Jenkins -> Configure System in the section E-mail Notifications fill in the current SMTP relay for OCP instances, which is smtp.ci.centos.org. Then, to check if everything works as it should, check the Test configuration by sending test e-mail checkbox, fill in the recipient, and click on Test configuration. If the email arrives, you should be all set.