So You Want to Contribute…¶
For general information on contributing to OpenStack, please check out the contributor guide to get started. It covers all the basics that are common to all OpenStack projects: the accounts you need, the basics of interacting with our Gerrit review system, how we communicate as a community, etc.
Below will cover the more project specific information you need to get started with placement.
As an official OpenStack project, Placement follows the overarching processes outlined in the Project Team Guide. Contribution is welcomed from any interested parties and takes many different forms.
To make sure everything gets the attention it deserves and work is not duplicated there are some guidelines, stated here.
If in doubt, ask someone, either by sending a message to the
openstack-discuss mailing list with a
[placement] subject tag, or by
#openstack-nova IRC channel on
Submitting and Managing Bugs¶
Bugs found in placement should be reported in Launchpad by creating a new
bug in the
If you are submitting a new bug, explain the problem, the steps taken that led to the bad results, and the expected results. Please also add as much of the following information as possible:
Relevant lines from the
The OpenStack release (e.g.,
The method used to install or deploy placement.
The operating system(s) on which the placement is running.
The version(s) of Python being used.
Tag the bug with
tags, like doc, api, etcetera.
Learn more about launchpad from openstack launchpad doc.
Triaging newly submitted bugs to confirm they really are bugs, gather missing information, and to suggest possible solutions is one of the most important contributions that can be made to any open source project.
If a new bug doesn’t have tags, add the relevant tag as per the area of affected code.
Leave comments on the bug if you have questions or ideas. If you are relatively certain about a solution, add the steps of that solution as tasks on the bug.
While triaging, only if you are sure, update the status of the bug from new to others.
If submitting a change related to a bug, the gerrit system will
automatically link to launchpad bug if you include
bug_id: identifiers in
your commit message, like this:
Related-Bug: 2005189 Partial-Bug: 2005190 Closes-Bug: 2005190
Like other OpenStack projects, Placement uses gerrit to facilitate peer code review. It is hoped and expected that anyone who would like to commit code to the Placement project will also spend time reviewing code for the sake of the common good. The more people reviewing, the more code that will eventually merge.
See How to Review Changes the OpenStack Way for an overview of the review and voting process.
There is a small group of people within the Placement team called core
reviewers. These are people who are empowered to signal (via the
that code is of a suitable standard to be merged and is aligned with the
current goals of the project. Core reviewers are regularly selected from all
active reviewers based on the quantity and quality of their reviews and
demonstrated understanding of the Placement code and goals of the project.
The point of review is to evolve potentially useful code to merged working code that is aligned with the standards of style, testing, and correctness that we share as group. It is not for creating perfect code. Review should always be constructive, encouraging, and friendly. People who contribute code are doing the project a favor, make it feel that way.
Some guidelines that reviewers and patch submitters should be aware of:
It is very important that a new patch set gets some form of review as soon as possible, even if only to say “we’ve seen this”. Latency in the review process has been identified as hugely discouraging for new and experienced contributors alike.
Follow up changes, to fix minor problems identified during review, are encouraged. We want to keep things moving.
As a reviewer, remember that not all patch submitters will know these guidelines. If it seems they don’t, point them here and be patient in the meantime.
Gerrit can be good for code review, but is often not a great environment for having a discussion that is struggling to resolve to a decision. Move discussion to the mailing list sooner rather than later. Add a link to the thread in the list archive to the review.
If the CI system is throwing up random failures in test runs, you should endeavor whenever possible to investigate, not simply
recheck. A flakey gate is an indication that OpenStack is not robust and at the root of all this, making OpenStack work well is what we are doing.
See here for How to Recheck
Special Considerations For Core Reviewers¶
Core reviewers have special powers. With great power comes great responsibility and thus being held to a standard. As a core reviewer, your job is to enable other people to contribute good code. Under ideal conditions it is more important to be reviewing other people’s code and bugs and fixing bugs than it is to be writing your own features. Frequently conditions will not be ideal, but strive to enable others.
When there are open questions that need to be resolved, try to prefer the openstack-discuss list over IRC so that anyone can be involved according to their own schedules and input from unexpected sources can be available.
This document cannot enumerate all the many ways to write good Python code. Instead it lists some guidelines that, if followed, will help make sure your code is reviewed promptly and merges quickly. As with everything else in this document, these guidelines will evolve over time and may be violated for special circumstances. If you have questions, ask.
See Placement Developer Notes for an overview of Placement and how the various pieces fit together.
Divide your change into a series of commits each of which encapsulates a single unit of functionality but still results in a working service. Smaller changes are easier to review.
If your change is to the HTTP API, familiarize yourself with Microversions.
Changes must include tests. There is a separate document on Testing Placement.
toxbefore submitting your code to gerrit. This will run unit and functional tests in both Python 2 and Python 3, and pep8 style checks. Placement tests, including functional, are fast, so this should not be too much of a hardship. By running the tests locally you avoid wasting scarce resources in the CI system.
Keep the tests fast. Avoid sleeps, network connections, and external processes in the tests.
Keep Placement fast. There is a
placement-perfloadjob that runs with every patch. Within that is a log file,
/logs/placement-perf.txt[.gz]that gives rough timing information for a common operation. We want those numbers to stay small.
We follow the code formatting guidelines of PEP 8. Check your code with
tox -epep8(for all files) or
tox -efast8(for just the files you changed). You will not always agree with the advice provided. Follow it.
Where possible avoid using the visual indent style. Using it can make future changes unnecessarily difficult. This guideline is not enforced by pep8 and has been used throughout the code in the past. There’s no need to fix old use. Instead of this
return_value = self.some_method(arg1, arg2, arg3, arg4)
return_value = self.some_method( arg1, arg2, arg3, arg4)
New functionality in Placement is developed as needed to meet new use cases or improve the handling of existing use cases. As a service used by other services in OpenStack, uses cases often originate in those other services. Considerable collaboration with other projects is often required to determine if any changes are needed in the Placement API or elsewhere in the project. That interaction should happen in the usual ways: At Project Team Gatherings, on the openstack-discuss list, and in IRC.
Create a new bug as described in New Bugs above.
If a spec is required there are some guidelines for creating one:
A file should be created in the placement code in
doc/source/specs/<cycle-name>/approvedwith a filename beginning with the identifier of the bug. For example:
More details on how to write a spec are included in a
template.rstfile found in the
doc/source/specsdirectory. This may be copied to use as the starting point for a new spec.
Under normal circumstances specs should be proposed near the beginning of a release cycle so there is sufficient time to review the spec and its implementation as well as to make any necessary decisions about limiting the number of specs being worked in the same cycle. Unless otherwise announced at the beginning of a cycle, specs should merge before milestone-2 to be considered relevant for that cycle. Exceptions will be reviewed on a case by case basis. See the stein schedule for an example schedule.
Work items that are described in a spec should be reflected as tasks created on the originating launchpad bug. Update the bug with additional tasks as they are discovered. Most new tasks will not require updating the spec.
If, when developing a feature, the implementation significantly diverges from the spec, the spec should be updated to reflect the new reality. This should not be considered exceptional: It is normal for there to be learning during the development process which impacts the solution.
Though specs are presented with the Placement documentation and can usefully augment end-user documentation, they are not a substitute. Development of a new feature is not complete without documentation.
When a spec was approved in a previous release cycle, but was not finished, it
should be re-proposed (via gerrit) to the current cycle. Include
Previously-Approved: <cycle> in the commit message to highlight that fact.
If there have been no changes, core reviewers should feel free to fast-approve
+2 required) the change.
Project Team Lead Duties¶
PTL duties are enumerated in the PTL guide.