rebase

.github/actions/ncp/action.yaml

@@ -3,10 +3,14 @@ description: Install ncp and display vm IP
 inputs:
   instance_id:
     required: false
-    description: "runner instance_id"
+    description: "runner instance-id"
   sa_json:
     required: true
     description: ""
+outputs:
+  runner_ipv4:
+    value: ${{ steps.result.outputs.runner_ipv4 }}
+    description: "runner ipv4"
 runs:
   using: composite
   steps:
@@ -22,8 +26,6 @@ runs:
     - name: configure ncp
       shell: bash
       run: |
-        set -x
-
         cat <<EOF > sa.json
         ${{ inputs.sa_json }}
         EOF
@@ -32,10 +34,12 @@ runs:
         ncp config set service-account-key sa.json
         ncp config set endpoint api.ai.nebius.cloud:443
 
-    - name: print ipv4 address of the vm
+    - id: result
+      name: print ipv4 address of the vm
       shell: bash
       run: |
         export GITHUB_RUNNER_IPV4=$(ncp compute instance get ${{ inputs.instance_id }} --format json | jq -r .network_interfaces[].primary_v4_address.one_to_one_nat.address)
         echo "GITHUB_RUNNER_IPV4=${GITHUB_RUNNER_IPV4}" >> $GITHUB_ENV
-        echo "ssh -i ~/.ssh/gh-runner github@${GITHUB_RUNNER_IPV4}"
+        echo "runner_ipv4=${GITHUB_RUNNER_IPV4}" >> $GITHUB_OUTPUT
+        echo "ssh -i ~/.ssh/gh-runner github@${GITHUB_RUNNER_IPV4}" | tee -a $GITHUB_SUMMARY
 

.github/actions/sleep/action.yaml

@@ -0,0 +1,20 @@
+name: Set sleep_after_tests variable
+description: Set up sleep after tests variable
+inputs:
+  sleep_after_tests:
+    required: false
+    default: '7200'
+    description: default sleep_after_tests value
+outputs:
+  sleep_after_tests:
+    value: ${{ steps.result.outputs.sleep_after_tests }}
+    description: "sleep_after_tests"
+runs:
+  using: composite
+  steps:
+    - id: result
+      name: set sleep_after_tests
+      shell: bash
+      run: |
+        echo "sleep_after_tests=${{inputs.sleep_after_tests}}" >> $GITHUB_OUTPUT
+

.github/actions/ssh_keys/action.yaml

@@ -0,0 +1,73 @@
+name: Add members ssh keys for VM
+description: Add team members public ssh keys to github@ user
+inputs:
+  org:
+    required: true
+    description: "org name"
+  team:
+    required: true
+    description: "team slug name"
+  token:
+    required: true
+    description: "GITHUB_TOKEN"
+
+runs:
+  using: composite
+  steps:
+    - name: init
+      shell: bash
+      run: |
+        echo "LOGINS_FILE=$(mktemp)" >> $GITHUB_ENV
+        echo "KEYS_FILE=$(mktemp)" >> $GITHUB_ENV
+    - name: collect members
+      shell: bash
+      if: ${{ github.repository == 'ydb-platform/nbs'}}
+      run: |
+        echo "https://api.github.com/orgs/${org}/teams/${team}/members?per_page=100&page=1"
+        curl -s -L -H "Accept: application/vnd.github+json" \
+             -H "Authorization: Bearer $token" \
+             -H "X-GitHub-Api-Version: 2022-11-28" \
+             "https://api.github.com/orgs/${org}/teams/${team}/members?per_page=100&page=1"
+        curl -s -L -H "Accept: application/vnd.github+json" \
+             -H "Authorization: Bearer $token" \
+             -H "X-GitHub-Api-Version: 2022-11-28" \
+             "https://api.github.com/orgs/${org}/teams/${team}/members?per_page=100&page=1" | jq -r .[].login | tee -a $LOGINS_FILE
+      env:
+        org: ${{ inputs.org }}
+        team: ${{ inputs.team }}
+        token: ${{ inputs.token }}
+    - name: collect members
+      shell: bash
+      if: ${{ github.repository != 'ydb-platform/nbs' }}
+      run: |
+        echo "https://api.github.com/orgs/${org}/members?per_page=100&page=1"
+        curl -s -L -H "Accept: application/vnd.github+json" \
+             -H "Authorization: Bearer $token" \
+             -H "X-GitHub-Api-Version: 2022-11-28" \
+             "https://api.github.com/orgs/${org}/members?per_page=100&page=1"
+        curl -s -L -H "Accept: application/vnd.github+json" \
+             -H "Authorization: Bearer $token" \
+             -H "X-GitHub-Api-Version: 2022-11-28" \
+             "https://api.github.com/orgs/${org}/members?per_page=100&page=1" | jq -r .[].login | tee -a $LOGINS_FILE
+      env:
+        org: ${{ inputs.org }}
+        team: ${{ inputs.team }}
+        token: ${{ inputs.token }}
+    - name: collect members public ssh keys
+      shell: bash
+      run: |
+        cat $LOGINS_FILE | while read login; do
+          curl  -s -L -H "Accept: application/vnd.github+json" \
+                -H "Authorization: Bearer $token" \
+                -H "X-GitHub-Api-Version: 2022-11-28" \
+                https://api.github.com/users/${login}/keys | jq -r .[].key | while read key; do
+                  echo $key $login;
+                done;
+        done | tee -a $KEYS_FILE
+      env:
+        token: ${{ inputs.token }}
+    - name: add ssh keys to user authorized_keys file
+      shell: bash
+      run: |
+        cat $KEYS_FILE >> /home/github/.ssh/authorized_keys
+

.github/actions/test/action.yaml

@@ -30,10 +30,6 @@ inputs:
   cache_update:
     required: false
     description: "Use cache for tests"
-  sleep_after_tests:
-    required: false
-    default: "0"
-    description: "Number of seconds to sleep after ya test launch"
 runs:
   using: composite
   steps:
@@ -120,8 +116,6 @@ runs:
             fi
         )
 
-        sleep ${{ inputs.sleep_after_tests }}
-
     - name: postprocess junit report
       shell: bash
       run: |

.github/workflows/build_and_test_cmake.yaml

@@ -13,6 +13,14 @@ on:
         type: string
         default: "linux"
         description: "runner label"
+      runner_instance_id:
+        type: string
+        default: ""
+        description: "runner instance id"
+      runner_ipv4:
+        type: string
+        default: ""
+        description: "runner ipv4"
       run_build:
         type: boolean
         default: true
@@ -21,12 +29,22 @@ on:
         type: boolean
         default: true
         description: "run tests"
+      sleep_after_tests:
+        type: string
+        default: "0"
+        description: "Amount of time to sleep after tests"
+    outputs:
+      sleep_after_tests:
+        description: "sleep_after_tests"
+        value: ${{ jobs.build-and-test.outputs.sleep_after_tests }}
 
 
 jobs:
   build-and-test:
-    name: Build and test ${{ inputs.build_preset }}
+    name: Build and test ${{ inputs.build_preset }} [id=${{ inputs.runner_instance_id }} ip=${{ inputs.runner_ipv4 }}]
     runs-on: [ self-hosted, "${{ inputs.runner_kind }}", "${{ inputs.runner_label }}" ]
+    outputs:
+      sleep_after_tests: ${{ steps.failure.outputs.sleep_after_tests }}
     timeout-minutes: 1440
     steps:
     - name: Checkout PR
@@ -60,4 +78,11 @@ jobs:
 
     - name: Run tests
       uses: ./.github/actions/test_cmake
-      if: inputs.run_tests
+      if: inputs.run_tests
+
+    - id: failure
+      name: set sleep_after_tests in case of failure
+      if: failure()
+      uses: ./.github/actions/sleep
+      with:
+        sleep_after_tests: ${{ inputs.sleep_after_tests != '0' && inputs.sleep_after_tests || '7200' }}

.github/workflows/build_and_test_on_demand.yaml

@@ -91,6 +91,7 @@ on:
         default: false
         description: "Update build cache in test stage"
 
+
 jobs:
   provide-runner:
     name: Start self-hosted runner
@@ -103,6 +104,7 @@ jobs:
       - name: Start YC runner
         id: start-yc-runner
         uses: librarian/[email protected]
+        timeout-minutes: 60
         with:
           mode: start
           yc-sa-json-credentials: ${{ secrets.NEBIUS_SA_JSON_CREDENTIALS }}
@@ -123,9 +125,11 @@ jobs:
           ssh-public-key: ${{ secrets.RUNNER_PUBLIC_KEY }}
 
   prepare-vm:
-    name: Prepare runner
+    name: Prepare runner [id=${{ needs.provide-runner.outputs.instance-id }}]
     needs: provide-runner
     runs-on: [ self-hosted, "${{ needs.provide-runner.outputs.label }}" ]
+    outputs:
+      runner_ipv4: ${{ steps.configure-ncp.outputs.runner_ipv4 }}
     steps:
       - name: Checkout PR
         uses: actions/checkout@v3
@@ -135,47 +139,62 @@ jobs:
       - name: Checkout
         uses: actions/checkout@v3
         if: github.event.pull_request.head.sha == ''
-      - name: Configure NCP
+      - id: configure-ncp
+        name: Configure NCP
         uses: ./.github/actions/ncp
         with:
           instance_id: ${{ needs.provide-runner.outputs.instance-id }}
           sa_json: ${{ secrets.NEBIUS_GITHUB_USER_SA_JSON}}
+      - id: configure-authorized-keys
+        name: Configure authorized_keys
+        uses: ./.github/actions/ssh_keys
+        with:
+          org: ydb-platform
+          team: nbs
+          token: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }}
       - name: Prepare VM
         uses: ./.github/actions/prepare
 
   build-and-test:
-    name: Build and test NBS
+    name: Build and test NBS [id=${{ needs.provide-runner.outputs.instance-id }} ip=${{ needs.prepare-vm.outputs.runner_ipv4 }}]
     uses: ./.github/workflows/build_and_test_ya.yaml
     needs:
       - provide-runner
       - prepare-vm
     with:
       runner_kind: self-hosted
       runner_label: ${{ needs.provide-runner.outputs.label }}
+      runner_instance_id: ${{ needs.provide-runner.outputs.instance-id }}
+      runner_ipv4: ${{ needs.prepare-vm.outputs.runner_ipv4 }}
       build_target: ${{ inputs.build_target }}
       test_target: ${{ inputs.test_target }}
       build_preset: ${{ inputs.build_preset }}
       run_build: ${{ inputs.run_build }}
       run_tests: ${{ inputs.run_tests }}
       test_size: ${{ inputs.test_size }}
       test_type: ${{ inputs.test_type }}
-      sleep_after_tests: ${{ inputs.sleep_after_tests }}
-      link_threads: 32
-      test_threads: 64
+      link_threads: 24
+      test_threads: 24
       cache_update_build: ${{ fromJson(github.event_name == 'workflow_dispatch' && format('{0}', inputs.cache_update) || format('{0}', inputs.cache_update_build)) }}
       cache_update_tests: ${{ fromJson(github.event_name == 'workflow_dispatch' && format('{0}', inputs.cache_update) || format('{0}', inputs.cache_update_tests)) }}
     secrets: inherit
 
   release-runner:
-    name: Release self-hosted runner
+    name: Release self-hosted runner ${{ needs.provide-runner.outputs.instance-id }}
     needs:
       - provide-runner # required to get output from the start-runner job
       - build-and-test # required to wait when the main job is done
     runs-on: ubuntu-latest
     if: always()
     steps:
+      - name: Sleep ${{ needs.build-and-test.outputs.sleep_after_tests }}s if build failed
+        shell: bash
+        if: ${{ needs.build-and-test.outputs.sleep_after_tests && needs.build-and-test.outputs.sleep_after_tests != '0'  }}
+        run: sleep ${{ needs.build-and-test.outputs.sleep_after_tests }}
+
       - name: Stop YC runner
         uses: librarian/[email protected]
+        timeout-minutes: 60
         with:
           mode: stop
           yc-sa-json-credentials: ${{ secrets.NEBIUS_SA_JSON_CREDENTIALS }}

.github/workflows/build_and_test_on_demand_cmake.yaml

@@ -20,6 +20,10 @@ on:
         type: boolean
         default: true
         description: "run tests"
+      sleep_after_tests:
+        type: string
+        default: "0"
+        description: "Amount of seconds to sleep after tests"
   workflow_call:
     inputs:
       build_preset:
@@ -31,6 +35,10 @@ on:
       run_tests:
         type: boolean
         default: true
+      sleep_after_tests:
+        type: string
+        default: "0"
+        description: "Amount of seconds to sleep after tests"
 
 jobs:
   provide-runner:
@@ -44,6 +52,7 @@ jobs:
       - name: Start YC runner
         id: start-yc-runner
         uses: librarian/[email protected]
+        timeout-minutes: 60
         with:
           mode: start
           yc-sa-json-credentials: ${{ secrets.NEBIUS_SA_JSON_CREDENTIALS }}
@@ -67,6 +76,8 @@ jobs:
     name: Prepare runner
     needs: provide-runner
     runs-on: [ self-hosted, "${{ needs.provide-runner.outputs.label }}" ]
+    outputs:
+      runner_ipv4: ${{ steps.configure-ncp.outputs.runner_ipv4 }}
     steps:
       - name: Checkout PR
         uses: actions/checkout@v3
@@ -76,26 +87,37 @@ jobs:
       - name: Checkout
         uses: actions/checkout@v3
         if: github.event.pull_request.head.sha == ''
-      - name: Configure NCP
+      - id: configure-ncp
+        name: Configure NCP
         uses: ./.github/actions/ncp
         with:
           instance_id: ${{ needs.provide-runner.outputs.instance-id }}
           sa_json: ${{ secrets.NEBIUS_GITHUB_USER_SA_JSON}}
+      - id: configure-authorized-keys
+        name: Configure authorized_keys
+        uses: ./.github/actions/ssh_keys
+        with:
+          org: ydb-platform
+          team: nbs
+          token: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }}
       - name: Prepare VM
         uses: ./.github/actions/prepare
 
   build-and-test:
-    name: Build and test NBS
+    name: Build and test NBS [id=${{ needs.provide-runner.outputs.instance-id }} ip=${{ needs.prepare-vm.outputs.runner_ipv4 }}]
     uses: ./.github/workflows/build_and_test_cmake.yaml
     needs:
       - provide-runner
       - prepare-vm
     with:
       runner_kind: self-hosted
       runner_label: ${{ needs.provide-runner.outputs.label }}
+      runner_instance_id: ${{ needs.provide-runner.outputs.instance-id }}
+      runner_ipv4: ${{ needs.prepare-vm.outputs.runner_ipv4 }}
       build_preset: ${{ inputs.build_preset }}
       run_build: ${{ inputs.run_build }}
       run_tests: ${{ inputs.run_tests }}
+      sleep_after_tests: ${{ inputs.sleep_after_tests }}
     secrets: inherit
 
   release-runner:
@@ -106,8 +128,14 @@ jobs:
     runs-on: ubuntu-latest
     if: always()
     steps:
+      - name: Sleep ${{ needs.build-and-test.outputs.sleep_after_tests || '0' }}s if build failed
+        shell: bash
+        if: ${{ needs.build-and-test.outputs.sleep_after_tests && needs.build-and-test.outputs.sleep_after_tests != '0'  }}
+        run: sleep ${{ needs.build-and-test.outputs.sleep_after_tests }}
+
       - name: Stop YC runner
         uses: librarian/[email protected]
+        timeout-minutes: 60
         with:
           mode: stop
           yc-sa-json-credentials: ${{ secrets.NEBIUS_SA_JSON_CREDENTIALS }}