Skip to content

fix(xds): eds deadlock on initial fetch timeout (backport of #11602) #7957

fix(xds): eds deadlock on initial fetch timeout (backport of #11602)

fix(xds): eds deadlock on initial fetch timeout (backport of #11602) #7957

name: "build-test-distribute"
on:
push:
branches: ["master", "release-*", "!*-merge-master"]
tags: ["*"]
pull_request:
branches: ["master", "release-*"]
concurrency:
group: ${{ github.head_ref || github.run_id }}
cancel-in-progress: true
permissions:
contents: read
env:
# This is automatically managed by CI
K8S_MIN_VERSION: v1.23.17-k3s1
K8S_MAX_VERSION: v1.28.1-k3s1
GH_OWNER: ${{ github.repository_owner }}
KUMA_DIR: "."
CI_TOOLS_DIR: /home/runner/work/kuma/kuma/.ci_tools
GH_USER: "github-actions[bot]"
GH_EMAIL: "<41898282+github-actions[bot]@users.noreply.github.com>"
jobs:
check:
timeout-minutes: 15
runs-on: ubuntu-latest
steps:
- name: "Fail when 'ci/force-publish' label is present on PRs from forks"
if: ${{ github.event_name == 'pull_request' && contains(github.event.pull_request.labels.*.name, 'ci/force-publish') && github.event.pull_request.head.repo.full_name != github.repository }}
run: |
echo "::error title=Label 'ci/force-publish' cannot be used on PRs from forks::To prevent accidental exposure of secrets, CI won't use repository secrets on pull requests from forks"
exit 1
- uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
with:
fetch-depth: 0
- uses: actions/setup-go@0c52d547c9bc32b1aa3301fd7a9cb496313a4491 # v5.0.0
with:
go-version-file: go.mod
cache: false
- uses: actions/cache@13aacd865c20de90d75de3b17ebe84f7a17d57d2 # v4.0.0
with:
path: |
${{ env.CI_TOOLS_DIR }}
key: ${{ runner.os }}-${{ runner.arch }}-devtools-${{ hashFiles('mk/dependencies/deps.lock') }}
restore-keys: |
${{ runner.os }}-${{ runner.arch }}-devtools
- run: |
make dev/tools
- uses: golangci/golangci-lint-action@3a919529898de77ec3da873e3063ca4b10e7f5cc # v3.7.0
with:
args: --fix=false --verbose
version: v1.55.2
skip-pkg-cache: true
- run: |
make clean
- run: |
make check
test:
runs-on: ubuntu-latest
if: ${{ ! contains(github.event.pull_request.labels.*.name, 'ci/skip-test') }}
steps:
- uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
with:
fetch-depth: 0
- uses: actions/setup-go@0c52d547c9bc32b1aa3301fd7a9cb496313a4491 # v5.0.0
with:
go-version-file: go.mod
- uses: actions/cache@13aacd865c20de90d75de3b17ebe84f7a17d57d2 # v4.0.0
with:
path: |
${{ env.CI_TOOLS_DIR }}
key: ${{ runner.os }}-${{ runner.arch }}-devtools-${{ hashFiles('mk/dependencies/deps.lock') }}
restore-keys: |
${{ runner.os }}-${{ runner.arch }}-devtools
- run: |
make dev/tools
- run: |
make test
distributions:
needs: ["check", "test", "test_e2e", "test_e2e_env"]
if: ${{ always() }}
runs-on: ubuntu-latest
steps:
- name: "Halt due to previous failures"
if: ${{ contains(needs.*.result, 'failure')|| contains(needs.*.result, 'cancelled') }}
run: |
exit 1
# for some reason, GH Action will always trigger a downstream job even if there are errors in an dependent job
# so we manually check it here. An example could be found here: https://github.com/kumahq/kuma/actions/runs/7044980149
- uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
with:
fetch-depth: 0
- name: "Maybe set full matrix"
if: github.event_name == 'push' || contains(github.event.pull_request.labels.*.name, 'ci/run-full-matrix')
id: set-full-matrix-switches
run: |
echo 'ENABLED_GOARCHES=arm64 amd64' >> $GITHUB_ENV
echo 'ENABLED_GOOSES=linux darwin' >> $GITHUB_ENV
- name: "Add matrix to .run-full-matrix for cache"
run: |
echo '${ENABLED_GOARCHES}|${ENABLED_GOOSES}' > .run-full-matrix
- name: "Maybe set flag to push build artifacts"
if: github.event_name == 'push' || contains(github.event.pull_request.labels.*.name, 'ci/force-publish')
run: |
echo 'ALLOW_PUSH=true' >> $GITHUB_ENV
- name: Install dependencies for cross builds
run: |
sudo apt-get update; sudo apt-get install -y qemu-user-static binfmt-support
- uses: actions/setup-go@0c52d547c9bc32b1aa3301fd7a9cb496313a4491 # v5.0.0
with:
go-version-file: go.mod
cache-dependency-path: |
.run-full-matrix
go.sum
- uses: actions/cache@13aacd865c20de90d75de3b17ebe84f7a17d57d2 # v4.0.0
with:
path: |
${{ env.CI_TOOLS_DIR }}
key: ${{ runner.os }}-${{ runner.arch }}-devtools-${{ hashFiles('mk/dependencies/deps.lock') }}
restore-keys: |
${{ runner.os }}-${{ runner.arch }}-devtools
- name: Free up disk space for the Runner
run: |
echo "Disk usage before cleanup"
sudo df -h
echo "Removing big directories"
sudo rm -rf /usr/share/dotnet /usr/local/lib/android /opt/ghc
echo "Removing images"
docker system prune --all -f
echo "Disk usage after cleanup"
sudo df -h
- run: |
make build
- run: |
make -j build/distributions
- run: |
make -j images
- run: |
make -j docker/save
- name: Run container structure test
if: ${{ !contains(github.event.pull_request.labels.*.name, 'ci/skip-container-structure-test') && !contains(github.event.pull_request.labels.*.name, 'ci/skip-test') }}
run: |
make test/container-structure
- name: Inspect created tars
run: |
for i in build/distributions/out/*.tar.gz; do echo $i; tar -tvf $i; done
- name: Publish distributions to Pulp
env:
PULP_USERNAME: ${{ vars.PULP_USERNAME }}
PULP_PASSWORD: ${{ secrets.PULP_PASSWORD }}
CLOUDSMITH_API_KEY: ${{ secrets.CLOUDSMITH_API_KEY }}
run: |
make publish/pulp
- name: Publish images
env:
DOCKER_API_KEY: ${{ secrets.DOCKER_API_KEY }}
DOCKER_USERNAME: ${{ vars.DOCKER_USERNAME }}
run: |-
make docker/login
# ensure we always logout
function on_exit() {
make docker/logout
}
trap on_exit EXIT
make docker/push
make docker/manifest
- name: package-helm-chart
id: package-helm
env:
HELM_DEV: ${{ !startsWith(github.event.ref, 'refs/tags/') }}
run: |
make helm/update-version
git config user.name "${GH_USER}"
git config user.email "${GH_EMAIL}"
git add -u deployments/charts
# This commit never ends up in the repo
git commit --allow-empty -m "ci(helm): update versions"
# To get an idea of what's in the commit to debug
git show
make helm/package
PKG_FILENAME=$(find .cr-release-packages -type f -printf "%f\n")
echo "filename=${PKG_FILENAME}" >> $GITHUB_OUTPUT
- name: Upload packaged chart
uses: actions/upload-artifact@694cdabd8bdb0f10b2cea11669e1bf5453eed0a6 # v4.2.0
with:
name: ${{ steps.package-helm.outputs.filename }}
path: .cr-release-packages/${{ steps.package-helm.outputs.filename }}
retention-days: ${{ github.event_name == 'pull_request' && 1 || 30 }}
# Everything from here is only running on releases.
# Ideally we'd finish the workflow early, but this isn't possible: https://github.com/actions/runner/issues/662
- name: Generate GitHub app token
id: github-app-token
if: ${{ startsWith(github.event.ref, 'refs/tags/') }}
uses: tibdex/github-app-token@3beb63f4bd073e61482598c45c71c1019b59b73a # v2.1.0
with:
app_id: ${{ secrets.APP_ID }}
private_key: ${{ secrets.APP_PRIVATE_KEY }}
- name: Release chart
if: ${{ startsWith(github.event.ref, 'refs/tags/') }}
env:
GITHUB_APP: "true"
GH_TOKEN: ${{ steps.github-app-token.outputs.token }}
run: make helm/release
gen_e2e_matrix:
runs-on: ubuntu-latest
if: ${{ !contains(github.event.pull_request.labels.*.name, 'ci/skip-test') && !contains(github.event.pull_request.labels.*.name, 'ci/skip-e2e-test') }}
outputs:
matrix: ${{ steps.generate-matrix.outputs.matrix }}
steps:
- id: generate-matrix
name: Generate matrix
env:
RUN_FULL_MATRIX: ${{ github.event_name == 'push' || contains(github.event.pull_request.labels.*.name, 'ci/run-full-matrix') }}
BASE_MATRIX: |-
{
"test_e2e": {
"target": [""],
"k8sVersion": ["kindIpv6", "${{ env.K8S_MIN_VERSION }}", "${{ env.K8S_MAX_VERSION }}"],
"arch": ["amd64"],
"parallelism": [3],
"cniNetworkPlugin": ["flannel"],
"legacyKDS": [false]
},
"test_e2e_env": {
"target": ["kubernetes", "universal", "multizone"],
"k8sVersion": ["kind", "kindIpv6", "${{ env.K8S_MIN_VERSION }}", "${{ env.K8S_MAX_VERSION }}"],
"arch": ["amd64"],
"parallelism": [1],
"cniNetworkPlugin": ["flannel"],
"legacyKDS": [false],
"exclude":[
{"target": "kubernetes", "k8sVersion":"kind"},
{"target": "multizone", "k8sVersion":"kind"},
{"target":"universal", "k8sVersion":"${{ env.K8S_MIN_VERSION }}"},
{"target":"universal", "k8sVersion":"${{ env.K8S_MAX_VERSION }}"}
],
"include":[
{"legacyKDS": true, "k8sVersion": "${{ env.K8S_MAX_VERSION }}", "target": "multizone", "arch": "amd64"},
{"k8sVersion": "${{ env.K8S_MAX_VERSION }}", "target": "multizone", "arch": "arm64"},
{"k8sVersion": "${{ env.K8S_MAX_VERSION }}", "target": "kubernetes", "arch": "arm64"},
{"k8sVersion": "${{ env.K8S_MAX_VERSION }}", "target": "universal", "arch": "arm64"},
{"k8sVersion": "${{ env.K8S_MAX_VERSION }}", "target": "gatewayapi", "arch": "amd64"},
{"cniNetworkPlugin": "calico", "k8sVersion": "${{ env.K8S_MAX_VERSION }}", "target": "multizone", "arch": "amd64"}
]
}
}
# You can modify the include to run one of test suites on PRs (though you'd need to then remove it)
OVERRIDE_JQ_CMD: |-
.test_e2e = false
| .test_e2e_env.include = []
| .test_e2e_env.exclude += [{"arch": "arm64"}, {"k8sVersion": "kindIpv6"}, {"k8sVersion": "${{ env.K8S_MIN_VERSION}}"}]
run: |-
BASE_MATRIX_ALL='${{ env.BASE_MATRIX }}'
if [[ "${{ env.RUN_FULL_MATRIX }}" != "true" ]]; then
BASE_MATRIX_ALL=$(echo $BASE_MATRIX_ALL | jq -r '${{ env.OVERRIDE_JQ_CMD }}')
fi
echo "final matrix: $BASE_MATRIX_ALL"
echo "matrix<<EOF" >> $GITHUB_OUTPUT
echo "$BASE_MATRIX_ALL" >> $GITHUB_OUTPUT
echo "EOF" >> $GITHUB_OUTPUT
test_e2e:
needs: ["gen_e2e_matrix"]
if: fromJSON(needs.gen_e2e_matrix.outputs.matrix).test_e2e
strategy:
matrix: ${{ fromJSON(needs.gen_e2e_matrix.outputs.matrix).test_e2e }}
fail-fast: false
uses: ./.github/workflows/e2e.yaml
with:
matrix: ${{ toJSON(matrix) }}
secrets:
circleCIToken: ${{ secrets.CIRCLECI_TOKEN }}
test_e2e_env:
needs: ["gen_e2e_matrix"]
if: fromJSON(needs.gen_e2e_matrix.outputs.matrix).test_e2e_env
strategy:
matrix: ${{ fromJSON(needs.gen_e2e_matrix.outputs.matrix).test_e2e_env }}
fail-fast: false
uses: ./.github/workflows/e2e.yaml
with:
matrix: ${{ toJSON(matrix) }}
secrets:
circleCIToken: ${{ secrets.CIRCLECI_TOKEN }}