The test-infra repository contains a collection of tools for testing Kubernetes and displaying Kubernetes tests results. There is no merge bot here. Feel free to click the big green merge button once your code is reviewed and passes on Travis. You will need to be a member of kubernetes/test-infra-maintainers to merge.
- The Kubernetes TestGrid shows historical test results.
- Configure your own testgrid dashboard at testgrid/config/config.yaml
- Gubernator formats the output of each run
- PR Dashboard finds PRs that need your attention
- Prow schedules testing and updates issues.
- Prow responds to GitHub events, timers and manual commands.
- The prow dashboard shows what it is currently testing
- Configure prow to run new tests at prow/config.yaml
- Triage Dashboard aggregates failures
- Triage clusters together similar failures
- Search for test failures across jobs
- Filter down to failures in a specific regex of tests and/or jobs.
- Test history is a deprecated tool
- Use the triage dashboard instead
- Summarizes the last 24 hours of testing.
- See Kettle and the corresponding bigquery metrics that largely supplement this information.
Test anything with the following pattern:
git clone https://github.com/kubernetes/test-infra
test-infra/jenkins/bootstrap.py --job=J --repo=R --service-account=S.json --upload=gs://B
The --job=J
flag specifies what test job to run.
The --repo=R
(or --bare
) flag controls what we check out from git.
Anyone can reconfigure our CI system with a test-infra PR that updates the appropriate files. Detailed instructions follow:
Create a PR in this repo to add/update/remove a job or suite. Specifically you'll need to do the following:
- Create an entry in
jobs/config.json
for the job- If this is a kubetest job create the corresponding
jobs/FOO.env
file - Ensure the
PROJECT=blah
in thejobs/FOO.env
has the right IAM grants
- If this is a kubetest job create the corresponding
- Add the job name to the
test_groups
list intestgrid/config/config.yaml
- Also the group to at least one
dashboard_tab
- Also the group to at least one
- Add the job to the appropriate section in
prow/config.yaml
- presubmit jobs run on unmerged code in PRs
- postsubmit jobs run after merging code
- periodic job run on a timed basis
- (Deprecated!) Some old jobs still run on jenkins
- Please do not add new jobs to jenkins
- Jenkins configuration is defined at
jenkins/job-configs
- More deprecated details at jenkins/README.md
Please test the job on your local workstation before creating a PR:
mkdir /tmp/whatever && cd /tmp/whatever
$GOPATH/src/k8s.io/test-infra/jenkins/bootstrap.py \
--job=J \ # aka your new job
--repo=R1 --repo=R2 \ # what repos to check out
--service-account ~/S.json # the service account to use to launch GCE/GKE clusters
# Note: create a service account at the cloud console for the project J uses
Presubmit will tell you if you forget to do any of this correctly.
Merge your PR and the job will be ready to go as soon as test-infra oncall
pushes the prow changes (make -C prow update-config
).
Largely similar to creating a new job, except you can just modify the existing entries rather than adding new ones.
Update what a job does by editing its definition in jobs/config.json
. For
the kubetest jobs this typically means editing the jobs/FOO.env
files it uses.
Update when a job runs by changing its definition in prow/config.yaml
.
The test-infra oncall must push prow changes (make -C prow update-config
).
Update where the job apears on testgrid by changing testgrid/config/config.yaml
.
The reverse of creating a new job: delete the appropriate entries in
jobs/config.json
, prow/config.yaml
and testgrid/config/config.yaml
.
The test-infra oncall must push prow changes (make -C prow update-config
).
We use Bazel to build and test the code in this repo.
The commands bazel build //...
and bazel test //...
should be all you need
for most cases. If you modify Go code, run ./verify/update-bazel.sh
to keep
BUILD
files up-to-date.
The Kubernetes project encourages organizations to contribute execution of e2e test jobs for a variety of platforms (e.g., Azure, rktnetes). The test-history scripts gather e2e results from these federated jobs. For information about how to contribute test results, see Federated Testing.