Skip to content

Node Throughput Test #51

Node Throughput Test

Node Throughput Test #51

name: Node Throughput Test
on:
schedule:
- cron: '39 0 * * *'
permissions:
# To be able to access the repository with actions/checkout
contents: read
concurrency:
# Structure:
# - Workflow name
# - Event type
# - A unique identifier depending on event type:
# - schedule: SHA
# - workflow_dispatch: PR number
#
# This structure ensures a unique concurrency group name is generated for each
# type of testing, such that re-runs will cancel the previous run.
group: |
${{ github.workflow }}
${{ github.event_name }}
${{
(github.event_name == 'schedule' && github.sha) ||
(github.event_name == 'workflow_dispatch' && github.event.inputs.PR-number)
}}
cancel-in-progress: true
env:
# renovate: datasource=github-releases depName=kubernetes/kops
kops_version: v1.28.1
# renovate: datasource=golang-version depName=go
go_version: 1.21.5
# Adding k8s.local to the end makes kops happy-
# has stricter DNS naming requirements.
test_name: node-throughput
cluster_base_name: ${{ github.run_id }}-${{ github.run_attempt }}.k8s.local
GCP_PERF_RESULTS_BUCKET: gs://cilium-scale-results
CILIUM_CLI_MODE: helm
jobs:
install-and-scaletest:
runs-on: ubuntu-latest
name: Install and Scale Test
timeout-minutes: 120
steps:
- name: Checkout context ref (trusted)
uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
with:
ref: ${{ inputs.context-ref || github.sha }}
persist-credentials: false
- name: Set Environment Variables
uses: ./.github/actions/set-env-variables
- name: Get Cilium's default values
id: default_vars
uses: ./.github/actions/helm-default
with:
image-tag: ${{ github.sha }}
- name: Set up job variables
id: vars
run: |
SHA="${{ github.sha }}"
# Setup Cilium install options
CILIUM_INSTALL_DEFAULTS="${{ steps.default_vars.outputs.cilium_install_defaults }} \
--wait=false"
CLUSTER_NAME="${{ env.test_name }}-${{ env.cluster_base_name }}"
echo SHA=${SHA} >> $GITHUB_OUTPUT
echo cilium_install_defaults=${CILIUM_INSTALL_DEFAULTS} >> $GITHUB_OUTPUT
echo CLUSTER_NAME=${CLUSTER_NAME} >> $GITHUB_OUTPUT
- name: Wait for images to be available
timeout-minutes: 30
shell: bash
run: |
for image in cilium-ci operator-generic-ci hubble-relay-ci ; do
until docker manifest inspect quay.io/${{ env.QUAY_ORGANIZATION_DEV }}/$image:${{ steps.vars.outputs.SHA }} &> /dev/null; do sleep 45s; done
done
- name: Install Go
uses: actions/setup-go@0c52d547c9bc32b1aa3301fd7a9cb496313a4491 # v5.0.0
with:
go-version: ${{ env.go_version }}
- name: Install Cilium CLI
uses: cilium/cilium-cli@c1fa8afda634f0457e9905d4f638e5164c4e0fc8 # v0.15.17
with:
repository: ${{ env.CILIUM_CLI_RELEASE_REPO }}
release-version: ${{ env.CILIUM_CLI_VERSION }}
- name: Install Kops
uses: cilium/scale-tests-action/install-kops@8a522b9d71254b6f6615c296e41a70610c9615ea # main
with:
release_version: ${{ env.kops_version }}
- name: Setup gcloud credentials
uses: google-github-actions/auth@67e9c72af6e0492df856527b474995862b7b6591 # v2.0.0
with:
credentials_json: '${{ secrets.GCP_PERF_SA_KEY }}'
- name: Setup gcloud CLI
uses: google-github-actions/setup-gcloud@825196879a077b7efa50db2e88409f44de4635c2 # v2.0.0
with:
project_id: ${{ secrets.GCP_PERF_PROJECT_ID }}
version: "405.0.0"
- name: Clone ClusterLoader2
uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
with:
repository: cilium/perf-tests
ref: 6f1bf283e22aa4c29d5db2a7b52501b091a0912e # master
persist-credentials: false
sparse-checkout: clusterloader2
path: perf-tests
- name: Display version info of installed tools
run: |
echo "--- go ---"
go version
echo "--- cilium-cli ---"
cilium version --client
echo "--- kops ---"
./kops version
echo "--- gcloud ---"
gcloud version
- name: Deploy cluster
id: deploy-cluster
uses: cilium/scale-tests-action/create-cluster@8a522b9d71254b6f6615c296e41a70610c9615ea # main
timeout-minutes: 30
with:
cluster_name: ${{ steps.vars.outputs.cluster_name }}
control_plane_size: n2-standard-4
control_plane_count: 1
node_size: e2-standard-8
node_count: 1
kops_state: ${{ secrets.GCP_PERF_KOPS_STATE_STORE }}
project_id: ${{ secrets.GCP_PERF_PROJECT_ID }}
- name: Setup firewall rules
uses: cilium/scale-tests-action/setup-firewall@8a522b9d71254b6f6615c296e41a70610c9615ea # main
with:
cluster_name: ${{ steps.vars.outputs.cluster_name }}
- name: Install Cilium
run: |
cilium install ${{ steps.vars.outputs.cilium_install_defaults }}
- name: Wait for cluster to be ready
uses: cilium/scale-tests-action/validate-cluster@8a522b9d71254b6f6615c296e41a70610c9615ea # main
timeout-minutes: 20
with:
cluster_name: ${{ steps.vars.outputs.cluster_name }}
kops_state: ${{ secrets.GCP_PERF_KOPS_STATE_STORE }}
- name: Wait for Cilium status to be ready
run: |
cilium status --wait
- name: Run CL2
id: run-cl2
working-directory: ./perf-tests/clusterloader2
timeout-minutes: 30
shell: bash
# --enable-exec-service=false to reduce number of pods so 100 pods can fit in node
# POD_STARTUP_LATENCY_THRESHOLD=60s so the test doesn't fail, currently we have ~30s pods startup latency
run: |
mkdir ./report
echo POD_STARTUP_LATENCY_THRESHOLD: 60s >> ./testoverrides.yaml
go run ./cmd/clusterloader.go \
-v=4 \
--testconfig=./testing/node-throughput/config.yaml \
--testoverrides=./testoverrides.yaml \
--enable-exec-service=false \
--provider=gce \
--enable-prometheus-server \
--tear-down-prometheus-server=false \
--report-dir=./report \
--kubeconfig=$HOME/.kube/config \
2>&1 | tee cl2-output.txt
- name: Get sysdump
if: ${{ always() && steps.run-cl2.outcome != 'skipped' }}
run: |
cilium status
cilium sysdump --output-filename cilium-sysdump-final
- name: Cleanup cluster
if: ${{ always() && steps.deploy-cluster.outcome != 'skipped' }}
uses: cilium/scale-tests-action/cleanup-cluster@8a522b9d71254b6f6615c296e41a70610c9615ea # main
with:
cluster_name: ${{ steps.vars.outputs.cluster_name }}
kops_state: ${{ secrets.GCP_PERF_KOPS_STATE_STORE }}
- name: Export results and sysdump to GS bucket
if: ${{ always() && steps.run-cl2.outcome != 'skipped' }}
uses: cilium/scale-tests-action/export-results@8a522b9d71254b6f6615c296e41a70610c9615ea # main
with:
test_name: ${{ env.test_name }}
results_bucket: ${{ env.GCP_PERF_RESULTS_BUCKET }}
artifacts: ./perf-tests/clusterloader2/report/*
other_files: cilium-sysdump-final.zip ./perf-tests/clusterloader2/cl2-output.txt