The infrastructure for the OpenStack project itself is run with the same processes, tools and philosophy as any other OpenStack project. The infrastructure team is an open meritocracy that welcomes new members. You can read about the OpenStack way on the wiki:
The project infrastructure encompasses all of the systems that are used in the day to day operation of the OpenStack project as a whole. This includes development, testing, and collaboration tools. All of the software that we run is open source, and its configuration is public. The project still uses a number of systems that do not yet fall under this umbrella (notably, the main website), but we’re working to incorporate them so that people may just as easily contribute to those areas. All new services used by the project should begin as part of the infrastructure project to ensure easy collaboration from the start.
We welcome contributions from new contributors. Reading this documentation is the first step. You should also join our mailing list.
We are most active on IRC, so please join the #openstack-infra channel on Freenode.
Feel free to attend our weekly IRC meeting. on Tuesdays at 19:00 UTC in #openstack-meeting.
Check out our open bugs on StoryBoard.
We hold regular bug days where we review and triage bugs.
To read about how our systems are managed and how to view or edit those configurations, see System Administration.
We also have a collection of OpenStack Project Infrastructure Publications where we host slides for presentations team members have given about the infrastructure.
And if you have any questions, please ask.
The infrastructure project maintains a bug list at:
Both defects and new features are tracked in the bug system. A number of tags are used to indicate relevance to a particular subsystem. There is also a low-hanging-fruit tag associated with bugs that should provide a gentle introduction to working on the infrastructure project without needing too much in-depth knowledge or access.
The infrastructure project designates a small number of efforts underway at any time as priority efforts. These are areas where the project has decided to focus resources to achieve major initiatives. These help reviewers prioritize their review workload and help to ensure the project accomplishes important tasks. Priority efforts are a great way to get involved in the project as they will generally provide the most interaction with experienced contributors.
Priority efforts are documented in the infra-specs repo. Each priority effort has one entry in infra-specs, though that may link to multiple smaller specifications for individual units of work if the effort is sufficiently large. Each priority effort also has a single person designated as the driver of that effort. That person is responsible for ensuring that anything blocking progress of the effort is discussed at team meetings and may be a good point of contact for someone who wants to get involved.
Changes not related to priority efforts will be reviewed by the core review team as time permits. This may mean that they spend considerable time in review, but they will be reviewed eventually. If a change is urgent, you might consider contacting someone in #openstack-infra on Freenode.
The infrastructure project is open, meaning anyone may join and begin contributing with no formal process. As an individual’s contributions and involvement grow, there are more formal roles. These roles are designed to empower groups of people to get work done in their area of expertise and interest, as well as supply a strong sense of direction for the infrastructure project as a whole. Everyone participating in the project is encouraged to expand their own knowledge while helping to support and mentor others as they progress.
The infrastructure project is composed of a large number of subprojects. Every source code repository has its own core team which is responsible for maintenance of that subproject, with some groups of repositories sharing a core team. These core teams are empowered to approve changes that reflect the currently understood project direction. Changes in project direction or major new initiatives must be approved by the infrastructure council.
Any existing core team member may nominate someone for addition to that core team by private communication with the infrastructure PTL. The PTL will consider the opinions of the existing core team members and the review history of the person in question, but final determination of core team membership (additions and removals) rests with the PTL. This process is private to enable honest evaluations in a safe environment.
Individuals who show an interest in a wide range of areas of the infrastructure project may be asked to join the infra-core team. To provide a baseline level of support to all of our subprojects and to ensure that important efforts may move forward, this team has approval rights in all infrastructure repositories. Members of this team may not be experts in all areas, but know their limits, and will not exceed those limits when reviewing changes outside of their area of expertise.
They are expected to have a wide general knowledge of what is going on in the infrastructure project and to help guide overall project direction. To that end, they are able to veto specs proposed to the infrastructure council.
The infrastructure council is the technical design body for the infrastructure project. While individuals and groups are empowered to execute the designs from the concil, major technical designs are agreed upon as a group to ensure that our large set of projects are all working together to the same end. The council need not delve too deeply into technical detail – just enough so that development efforts may happen in parallel and work toward a common goal.
All members of any infrastructure project core team have a seat on the Council. The Council is responsible for approving changes in project direction, major new initiatives, setting priority efforts, and addition or removal of projects.
Any such changes should be proposed to the infra-specs repository. Anyone is welcome to review specs and they are expected to evolve during the review process. When a change to infra-specs is ready for final approval, the author will add the change to the infra team meeting agenda so that the entire team is notified that the spec is ready. Members of the council will vote by leaving reviews on the spec to approve or reject the change. The determination will be based on a majority vote, with members of the infra-core team able to veto, and in the case of a tie, the PTL will cast the deciding vote. The PTL will execute the workflow action on the change after the vote.
Specs are living documents, and once approved, should be maintained for the duration of the effort they describe. Substantial changes in direction should go through the same process described above. The PTL may chose to approve insubstantial changes without the formal council voting process.
While core membership is analogous to the same system in other OpenStack projects, because the infrastructure team operates production servers there is another sub-group of the infrastructure team that has root access to all servers. Root membership is handled in the same way as core membership. Root members must also be infra-core members, but infra-core members may not necessarily be root members. This is because primary system administration is performed through code review, so anyone able to log into a machine to execute commands must be able to approve those same commands in configuration management; otherwise it would be easier for a person to bypass puppet than use it in the intended fashion.
Root access is generally only necessary to launch new servers, perform low-level maintenance, manage DNS, or fix problems. In general it is not needed for day-to-day system administration and configuration which is done in puppet (where anyone may propose changes). Therefore it is generally reserved for people who are well versed in infrastructure operations and can commit to spending a significant amount of time troubleshooting on servers.
Some individuals may need root access to individual servers; in these cases the infra-core group may grant root access on a limited basis.
We review each others changes before they are merged. This helps us improve the quality of the code we produce as well as ensure that we are working together as a team. Generally we expect at least two members of the core review team to approve a change before it is merged, but we are flexible in this requirement – typo fixes, or other simple changes may be approved with less formality.
The primary purpose of change review is to catch substantial errors before they are merged. In order to keep this process useful and avoid frustration for both authors and reviewers, please do not leave negative reviews for insubstantial faults or potential improvements. The purpose is not to make someone else’s code match your vision of perfection, but to enable all of us to work together on a project.
Please use discretion when deciding what is important enough for someone to spend the time to rework and for you to spend the time re-reviewing. Sometimes minor things are important, such as consistent use of hyphens versus underscores in a configuration language. Sometimes they are not, such as whitespace in documentation.
If you would like to mention minor improvements such as this, feel free to do so, but please do not leave a negative score on the review. If you mention them along with other more substantial criticisms, please note them in a review, for example, with “(nit)” or “(not a -1)” or “you may want to fix this if you are updating the patch anyway”.
Please also see the section in the Infrastructure Manual on peer review.