- Access to the fedora-coreos-pipeline namespace in the Fedora OpenShift instance
The steps for doing a release are in the tickets created in the coreos/fedora-coreos-streams repository. An overview of the process can be seen in this video.
Those issues are generally created at the end of the previous release but if you need to manually create one, you can use the following links:
If a part of a pipeline run fails there may be options available that don't require you to start from the very beginning.
- fedora-coreos-pipeline
- if the top level pipeline fails you'll need to restart it likely with FORCE checked in the build parameters.
- multi-arch-pipeline
- Assuming the top level pipeline passed you can start another multi-arch
pipeline run with the exact same parameters as the original multi-arch
pipeline run that was automatically kicked off by top level pipeline.
- Open the failed job and look at the build parameters that were used.
- Start a new job copy and pasting the exact build parameters.
- Assuming the top level pipeline passed you can start another multi-arch
pipeline run with the exact same parameters as the original multi-arch
pipeline run that was automatically kicked off by top level pipeline.
- kola-{aws,gcp,openstack}
- These jobs are just tests, that don't produce any artifacts.
- If one of these jobs fail you can start a new job, but make sure to use the exact same build parameters as the original failed job.
FCOS release managers are not part of the normal release rotation but pay attention to most releases and make sure they are progressing (PR reviews, debugging issues). The current release managers are:
- Dusty Mabe
- Jonathan Lebon
Release executors perform scheduled biweekly releases. The current set of release executors are (in alphabetical order, see below for the rotation):
- Adam Piasecki
- Michael Armijo
- Steven Presti
This list should be kept to three or four people so each executor performs releases frequently. If the list is any larger, it's too easy to lose familiarity with the process. When a new executor is added, the most experienced release executor will typically move to the standby list below.
The build pipeline for releases will typically get started on Monday. Assuming things go smoothly rollouts will start on Tuesday. If there are any conflicts in the schedule we can easily swap a rotation with someone else.
Standby release executors perform any release outside of the normal biweekly schedule. The current set of standby executors are (in alphabetical order, see below for the rotation):
- Clément Verna
- Gursewak Singh
- Renata Ravanelli
- Timothée Ravier
Out-of-cycle releases are often scheduled on short notice, time-sensitive, and/or require workarounds for bugs or infrastructure problems. All standby executors should be experienced release executors who are comfortable with the release process.
See the schedule in https://hackmd.io/WCA8XqAoRvafnja01JG_YA.