Skip to content

koord-descheduler: enhance LowNodeLoad scorer #204

koord-descheduler: enhance LowNodeLoad scorer

koord-descheduler: enhance LowNodeLoad scorer #204

Workflow file for this run

name: E2E-K8S-1.28
on:
push:
branches:
- main
- release-*
pull_request: {}
workflow_dispatch: {}
env:
# Common versions
GO_VERSION: '1.20'
KIND_ACTION_VERSION: 'v1.5.0'
KIND_VERSION: 'v0.22.0'
KIND_IMAGE: 'kindest/node:v1.28.7'
KIND_CLUSTER_NAME: 'ci-testing'
COMPONENT_NS: "koordinator-system"
jobs:
slo-controller:
continue-on-error: true
runs-on: ubuntu-20.04
steps:
- uses: actions/checkout@v4
with:
submodules: true
- name: Setup Go
uses: actions/setup-go@v5
with:
go-version: ${{ env.GO_VERSION }}
- name: Setup Kind Cluster
uses: helm/kind-action@v1.9.0
with:
node_image: ${{ env.KIND_IMAGE }}
cluster_name: ${{ env.KIND_CLUSTER_NAME }}
config: ./test/kind-conf.yaml
version: ${{ env.KIND_VERSION }}
- name: Build image
run: |
export MANAGER_IMAGE="koordinator-sh/koord-manager:e2e-${GITHUB_RUN_ID}"
docker build --pull . -t ${MANAGER_IMAGE} -f docker/koord-manager.dockerfile
export KOORDLET_IMAGE="koordinator-sh/koordlet:e2e-${GITHUB_RUN_ID}"
docker build --pull . -t ${KOORDLET_IMAGE} -f docker/koordlet.dockerfile
export SCHEDULER_IMAGE="koordinator-sh/koord-scheduler:e2e-${GITHUB_RUN_ID}"
docker build --pull . -t ${SCHEDULER_IMAGE} -f docker/koord-scheduler.dockerfile
kind load docker-image --name=${KIND_CLUSTER_NAME} ${MANAGER_IMAGE} || { echo >&2 "kind not installed or error loading image: ${MANAGER_IMAGE}"; exit 1; }
kind load docker-image --name=${KIND_CLUSTER_NAME} ${KOORDLET_IMAGE} || { echo >&2 "kind not installed or error loading image: ${KOORDLET_IMAGE}"; exit 1; }
kind load docker-image --name=${KIND_CLUSTER_NAME} ${SCHEDULER_IMAGE} || { echo >&2 "kind not installed or error loading image: ${SCHEDULER_IMAGE}"; exit 1; }
- name: Check host environment
run: |
set -ex
kubectl version --short
kubectl get pods -A
kubectl get nodes -o yaml
tree -L 2 /sys/
tree -L 2 /sys/fs/cgroup
cat /proc/cpuinfo
- name: Install Koordinator
run: |
set -ex
kubectl cluster-info
MANAGER_IMG=koordinator-sh/koord-manager:e2e-${GITHUB_RUN_ID} KOORDLET_IMG=koordinator-sh/koordlet:e2e-${GITHUB_RUN_ID} SCHEDULER_IMG=koordinator-sh/koord-scheduler:e2e-${GITHUB_RUN_ID} ./hack/deploy_kind.sh
NODES=$(kubectl get node | wc -l)
for ((i=1;i<10;i++));
do
set +e
PODS=$(kubectl get pod -n ${COMPONENT_NS} | grep "koord-manager\|koordlet" | grep '1/1' | wc -l)
set -e
if [ "$PODS" -ge "$NODES" ]; then
break
fi
sleep 6
done
set +e
PODS=$(kubectl get pod -n ${COMPONENT_NS} | grep "koord-manager\|koordlet" | grep '1/1' | wc -l)
kubectl get pod -A
kubectl get node -o yaml
kubectl get all -n ${COMPONENT_NS} -o wide
kubectl get pod -n ${COMPONENT_NS} --no-headers | grep koord-manager | head -n 1 | awk '{print $1}' | xargs kubectl logs -n ${COMPONENT_NS} --tail=100
kubectl get pod -n ${COMPONENT_NS} --no-headers | grep koord-scheduler | awk '{print $1}' | xargs kubectl logs -n ${COMPONENT_NS} --tail=100
kubectl get pod -n ${COMPONENT_NS} --no-headers | grep koordlet | head -n 1 | awk '{print $1}' | xargs -L 1 kubectl logs -n ${COMPONENT_NS}
kubectl get pod -n ${COMPONENT_NS} -o wide
set -e
if [ "$PODS" -ge "$NODES" ]; then
echo "Wait for koord-manager and koordlet ready successfully"
else
echo "Timeout to wait for koord-manager and koordlet ready"
exit 1
fi
- name: Run E2E Tests
run: |
export KUBECONFIG=/home/runner/.kube/config
make ginkgo
set +e
EXTRA_ARGS="-koordinator-component-namespace=${COMPONENT_NS} -allowed-not-ready-nodes=1 -system-pods-startup-timeout=10s -e2e-verify-service-account=false"
./bin/ginkgo -timeout 60m -v --focus='slo-controller' test/e2e -- ${EXTRA_ARGS}
retVal=$?
restartCount=$(kubectl get pod -n ${COMPONENT_NS} -l koord-app=koord-manager --no-headers | head -n 1 | awk '{print $4}')
if [ "${restartCount}" -eq "0" ];then
echo "koord-manager has not restarted"
else
kubectl get pod -n ${COMPONENT_NS} -l koord-app=koord-manager --no-headers
echo "koord-manager has restarted, abort!!!"
kubectl get pod -n ${COMPONENT_NS} --no-headers -l koord-app=koord-manager | head -n 1 | awk '{print $1}' | xargs kubectl logs -p -n ${COMPONENT_NS}
exit 1
fi
exit $retVal