Update GKE job to support the new cloud credentials (#5946)

* Add a new credential job to generate temporary credentials.
* Add a new job for ccredential renewal.
* GKE job is now triggered as a post-build action of the credential job.
* Update the GKE script to support the new login method.

Signed-off-by: Shuyang Xin <[email protected]>

ci/jenkins/jobs/job-templates.yaml

@@ -284,6 +284,30 @@
      publishers: '{publishers}'
      wrappers: '{wrappers}'
 
+- job-template:
+     name: 'cloud-{name}-{test_name}-credential'
+     node: '{node}'
+     triggers: '{triggers}'
+     builders: '{builders}'
+     parameters:
+     - string:
+         default: ''
+         description: The cloud token ID of the cloud provider.
+         name: ID_TOKEN
+         trim: 'false'
+     description: '{description}'
+     block-downstream: false
+     block-upstream: false
+     project-type: freestyle
+     properties:
+     - build-discarder:
+         artifact-days-to-keep: -1
+         artifact-num-to-keep: -1
+         days-to-keep: 30
+         num-to-keep: 30
+     publishers: '{publishers}'
+     wrappers: '{wrappers}'
+
 - job-template:
      name: 'cloud-{name}-{test_name}-cleanup'
      node: '{node}'

ci/jenkins/jobs/projects-cloud.yaml

@@ -718,10 +718,22 @@
           builders:
           - shell: |-
               #!/bin/bash
-              # gcloud auth login requires verification code from url
-              ${{GCLOUD_SDK_PATH}}/bin/gcloud auth activate-service-account --key-file=${{GCLOUD_KEY_PATH}}
+              # Prevent exposing credentials in the console output by adding set +x
+              # This is to avoid other developers removing this set+x by accident
+              set +x
+              token_file="/var/lib/jenkins/gcp_cred"
+              cp $token_file ${{WORKSPACE}}/gcp_cred
+              sudo gcloud iam workforce-pools create-cred-config \
+                locations/global/workforcePools/${{WORKFORCE_POOL}}/providers/oidc-provider \
+                --subject-token-type urn:ietf:params:oauth:token-type:id_token \
+                --credential-source-file gcp_cred \
+                --workforce-pool-user-project antrea\
+                --project antrea\
+                --output-file gcp_cred_config.json
+              sudo gcloud auth login --cred-file=gcp_cred_config.json
+              sudo gcloud config set project antrea
               sudo ./ci/test-conformance-gke.sh --cluster-name antrea-gke-${{BUILD_NUMBER}} \
-                --svc-account antrea-gcp@antrea.iam.gserviceaccount.com --gcloud-sdk-path ${{GCLOUD_SDK_PATH}} \
+                --skip-iam-policy-binding --gcloud-sdk-path ${{GCLOUD_SDK_PATH}} \
                 --log-mode detail --setup-only
           triggers:
           - timed: H H */2 * *
@@ -741,6 +753,86 @@
           - email:
               notify-every-unstable-build: true
               recipients: [email protected]
+          wrappers:
+          - credentials-binding:
+            - text:
+                credential-id: WORKFORCE_POOL # Jenkins secret that stores the cloud resource pool id
+                variable: WORKFORCE_POOL
+      - 'cloud-{name}-{test_name}-credential':
+          test_name: gke
+          node: antrea-credential
+          description: 'This is the cloud credential job for antrea gke tests.'
+          builders:
+          - shell: |-
+              #!/bin/bash
+              # Prevent exposing credentials in the console output by adding set +x
+              # This is to avoid other developers removing this set+x by accident
+              set +x
+              token_file="/var/lib/jenkins/ci_properties.txt"
+              if [ -e "$token_file" ]; then
+                  modification_timestamp=$(stat -c %Y "$token_file")
+                  current_timestamp=$(date +%s)
+                  time_difference=$((current_timestamp - modification_timestamp))
+              else
+                  time_difference=14401
+              fi
+              # The credentials are valid for 12 hours, to allow sufficient time for job execution, the reuse threshold is set to 4 hours.
+              threshold=14400
+              if [ "$time_difference" -gt "$threshold" ]; then
+                  echo "Generate the new Cloud Credential"
+                  json_data=$(python3 get_access_using_api_client.py ${{CLOUD_CLIENT_ID}} ${{CLOUD_CLIENT_TOKEN}} PowerUser gcp ${{GKE_RESOURCE_ID}} 43200 prd)
+                  json_data=$(echo $json_data | sed "s/'/\"/g" | sed 's/True/true/')
+                  id_token=$(echo $json_data | jq -r '.credential.idToken')
+                  echo "ID_TOKEN=$id_token" > ${{WORKSPACE}}/ci_properties.txt
+                  cp ${{WORKSPACE}}/ci_properties.txt $token_file
+              else
+                  echo "Reuse the unexpired Cloud Credential"
+                  cp $token_file ${{WORKSPACE}}/ci_properties.txt
+              fi
+          triggers:
+          - timed: H H/6 * * *
+          publishers:
+          - trigger-parameterized-builds:
+            - project:
+              - cloud-{name}-renew-credential
+              current-parameters: true
+              property-file: 'ci_properties.txt'
+          - email:
+              notify-every-unstable-build: true
+              recipients: [email protected]
+          wrappers:
+          - credentials-binding:
+            - text:
+                credential-id: CLOUD_CLIENT_ID # Jenkins secret that stores client id
+                variable: CLOUD_CLIENT_ID
+            - text:
+                credential-id: CLOUD_CLIENT_TOKEN # Jenkins secret that stores client secret token
+                variable: CLOUD_CLIENT_TOKEN
+            - text:
+                credential-id: GKE_RESOURCE_ID
+                variable: GKE_RESOURCE_ID
+      - 'cloud-{name}-{test_name}-credential':
+          test_name: renew
+          node: antrea-cloud
+          description: 'This is a periodic job to renew the credential on cloud node.'
+          builders:
+          - shell: |-
+              #!/bin/bash
+              # Prevent exposing credentials in the console output by adding set +x
+              # This is to avoid other developers removing this set+x by accident
+              set +x
+              cd /var/lib/jenkins
+              if [ -z "$ID_TOKEN" ]; then
+                  echo "No Token ID Found"
+                  exit 1
+              else
+                  echo "${{ID_TOKEN}}" > gcp_cred
+              fi
+          publishers:
+          - email:
+              notify-every-unstable-build: true
+              recipients: [email protected]
+          triggers: []
           wrappers: []
       - 'cloud-{name}-{test_name}-cleanup':
           test_name: gke

ci/test-conformance-gke.sh

@@ -29,6 +29,7 @@ MODE="report"
 RUN_ALL=true
 RUN_SETUP_ONLY=false
 RUN_CLEANUP_ONLY=false
+SKIP_IAM_POLICY_BINDING=false
 TEST_SCRIPT_RC=0
 KUBE_CONFORMANCE_IMAGE_VERSION=auto
 
@@ -89,6 +90,10 @@ case $key in
     USER_EMAIL="$2"
     shift 2
     ;;
+    --skip-iam-policy-binding)
+    SKIP_IAM_POLICY_BINDING=true
+    shift
+    ;;
     --gke-zone)
     GKE_ZONE="$2"
     shift 2
@@ -244,6 +249,8 @@ function deliver_antrea_to_gke() {
     elif [[ -n ${USER_EMAIL+x} ]]; then
         gcloud projects add-iam-policy-binding ${GKE_PROJECT} --member user:${USER_EMAIL} --role roles/container.admin
         kubectl create clusterrolebinding cluster-admin-binding --clusterrole cluster-admin --user ${USER_EMAIL}
+    elif [[ "$SKIP_IAM_POLICY_BINDING" == true ]]; then
+        echo "Skipping the IAM Policy Binding for Cluster Management."
     else
         echo "Neither service account or user email info is set, cannot create cluster-admin-binding!"
         echo "Please refer to --help for more information."