One ci.centos.org jenkins job that:
- Builds Che
- Publishes
che-server
on Docker registries (dockerhub and registry.devshift.net/che/che) - Deploys Che to dev.rdu2c.fabric8.io and [che.ci.centos.org] (https://che.ci.centos.org:8443)
One fabric8.io jenkins job that is triggered by step 2 of ci.centos.org job. This job is responsible for the release of a new version of Che available for openshift.io.
- Step 3. is just a deployment of Che but we want to add more tests to validate our builds: ideally an end to end test run on openshift.io. That is something which we desperately need and if we can't have an e2e test an
oc deploy
followed by acurl
to start a che workspace would be enought now. - fabric8.io build should be triggered only if validation is successfull (i.e. after step 3): this means that in step 2. we should publish images to a "stage" dockerhub repository (i.e rh-che-stage/che-server) and that we should add a fourth step to publish promoted images only to rh-che/che-server.
- Currently step 2. pushes only one tag of rh-che/che-server. We need to push more tags (without-keycloak, nightly etc...). That is trivial if we use a "stage" dockerhub repository to push these tags.
- We want use the same job to validate Pull Requests. In the case of PR validation:
- fabric8.io build should never be triggered (i.e. image should not be published to rh-che/che-server)
- the branch eclipse/che and redhat-developer/rh-che branches should be parameters of the jenkins job
- We don't need ci.centos.org build to be triggered twice a day. Trigger once a commit is merged in openshift-connector branch (for eclipse/che) or master branch (for redhat-developer/rh-che) is enough.