Revert "ci: Drop docker tests"

This reverts commit d82eb8d0f1.

.github/workflows/basic-ci-amd64.yaml

@@ -35,7 +35,7 @@ jobs:
       KATA_HYPERVISOR: ${{ matrix.vmm }}
       SANDBOXER: "shim"
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0
@@ -96,7 +96,7 @@ jobs:
       KATA_HYPERVISOR: ${{ matrix.vmm }}
       SANDBOXER: "podsandbox"
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0
@@ -141,7 +141,7 @@ jobs:
       GOPATH: ${{ github.workspace }}
       KATA_HYPERVISOR: ${{ matrix.vmm }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0
@@ -195,7 +195,7 @@ jobs:
     env:
       KATA_HYPERVISOR: ${{ matrix.vmm }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0
@@ -242,7 +242,7 @@ jobs:
       GOPATH: ${{ github.workspace }}
       KATA_HYPERVISOR: ${{ matrix.vmm }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0
@@ -269,6 +269,50 @@ jobs:
         timeout-minutes: 15
         run: bash tests/functional/vfio/gha-run.sh run
 
+  run-docker-tests:
+    name: run-docker-tests
+    strategy:
+      # We can set this to true whenever we're 100% sure that
+      # all the tests are not flaky, otherwise we'll fail them
+      # all due to a single flaky instance.
+      fail-fast: false
+      matrix:
+        vmm:
+          - qemu
+    runs-on: ubuntu-22.04
+    env:
+      KATA_HYPERVISOR: ${{ matrix.vmm }}
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          ref: ${{ inputs.commit-hash }}
+          fetch-depth: 0
+          persist-credentials: false
+
+      - name: Rebase atop of the latest target branch
+        run: |
+          ./tests/git-helper.sh "rebase-atop-of-the-latest-target-branch"
+        env:
+          TARGET_BRANCH: ${{ inputs.target-branch }}
+
+      - name: Install dependencies
+        run: bash tests/integration/docker/gha-run.sh install-dependencies
+        env:
+          GH_TOKEN: ${{ github.token }}
+
+      - name: get-kata-tarball
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4.3.0
+        with:
+          name: kata-static-tarball-amd64${{ inputs.tarball-suffix }}
+          path: kata-artifacts
+
+      - name: Install kata
+        run: bash tests/integration/docker/gha-run.sh install-kata kata-artifacts
+
+      - name: Run docker smoke test
+        timeout-minutes: 5
+        run: bash tests/integration/docker/gha-run.sh run
+
   run-nerdctl-tests:
     name: run-nerdctl-tests
     strategy:
@@ -287,7 +331,7 @@ jobs:
     env:
       KATA_HYPERVISOR: ${{ matrix.vmm }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0
@@ -334,7 +378,7 @@ jobs:
     name: run-kata-agent-apis
     runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/basic-ci-s390x.yaml

@@ -35,7 +35,7 @@ jobs:
       KATA_HYPERVISOR: ${{ matrix.vmm }}
       SANDBOXER: "shim"
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0
@@ -96,7 +96,7 @@ jobs:
       KATA_HYPERVISOR: ${{ matrix.vmm }}
       SANDBOXER: "podsandbox"
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0
@@ -123,3 +123,44 @@ jobs:
       - name: Run containerd-stability tests
         timeout-minutes: 15
         run: bash tests/stability/gha-run.sh run
+
+  run-docker-tests:
+    name: run-docker-tests
+    strategy:
+      # We can set this to true whenever we're 100% sure that
+      # all the tests are not flaky, otherwise we'll fail them
+      # all due to a single flaky instance.
+      fail-fast: false
+      matrix:
+        vmm: ['qemu']
+    runs-on: s390x-large
+    env:
+      KATA_HYPERVISOR: ${{ matrix.vmm }}
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          ref: ${{ inputs.commit-hash }}
+          fetch-depth: 0
+          persist-credentials: false
+
+      - name: Rebase atop of the latest target branch
+        run: |
+          ./tests/git-helper.sh "rebase-atop-of-the-latest-target-branch"
+        env:
+          TARGET_BRANCH: ${{ inputs.target-branch }}
+
+      - name: Install dependencies
+        run: bash tests/integration/docker/gha-run.sh install-dependencies
+
+      - name: get-kata-tarball
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4.3.0
+        with:
+          name: kata-static-tarball-s390x${{ inputs.tarball-suffix }}
+          path: kata-artifacts
+
+      - name: Install kata
+        run: bash tests/integration/docker/gha-run.sh install-kata kata-artifacts
+
+      - name: Run docker smoke test
+        timeout-minutes: 5
+        run: bash tests/integration/docker/gha-run.sh run

.github/workflows/build-checks-preview-riscv64.yaml

@@ -72,7 +72,7 @@ jobs:
           sudo rm -f /tmp/kata_hybrid*  # Sometime we got leftover from test_setup_hvsock_failed()
 
       - name: Checkout the code
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           fetch-depth: 0
           persist-credentials: false

.github/workflows/build-checks.yaml

@@ -84,7 +84,7 @@ jobs:
           sudo rm -f /tmp/kata_hybrid*  # Sometime we got leftover from test_setup_hvsock_failed()
 
       - name: Checkout the code
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           fetch-depth: 0
           persist-credentials: false

.github/workflows/build-kata-static-tarball-amd64.yaml

@@ -75,7 +75,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history
@@ -178,7 +178,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history
@@ -272,7 +272,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history
@@ -326,7 +326,7 @@ jobs:
       contents: read
       packages: write
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0
@@ -391,7 +391,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history
@@ -436,7 +436,7 @@ jobs:
       contents: read
       packages: write
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/build-kata-static-tarball-arm64.yaml

@@ -64,7 +64,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history
@@ -162,7 +162,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history
@@ -253,7 +253,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history
@@ -305,7 +305,7 @@ jobs:
       contents: read
       packages: write
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/build-kata-static-tarball-ppc64le.yaml

@@ -51,7 +51,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history
@@ -109,7 +109,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history
@@ -184,7 +184,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history
@@ -240,7 +240,7 @@ jobs:
         run: |
           sudo chown -R "$USER":"$USER" "$GITHUB_WORKSPACE"
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/build-kata-static-tarball-riscv64.yaml

@@ -38,7 +38,7 @@ jobs:
           - kernel
           - virtiofsd
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history

.github/workflows/build-kata-static-tarball-s390x.yaml

@@ -58,7 +58,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history
@@ -143,7 +143,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history
@@ -196,7 +196,7 @@ jobs:
       contents: read
       packages: write
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
       - name: Rebase atop of the latest target branch
@@ -270,7 +270,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0 # This is needed in order to keep the commit ids history
@@ -328,7 +328,7 @@ jobs:
       contents: read
       packages: write
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/build-kubectl-image.yaml

@@ -28,7 +28,7 @@ jobs:
       packages: write
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
 

.github/workflows/cargo-deny-runner.yaml

@@ -20,7 +20,7 @@ jobs:
 
     steps:
       - name: Checkout Code
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
       - name: Generate Action

.github/workflows/ci-weekly.yaml

@@ -73,7 +73,7 @@ jobs:
     runs-on: ubuntu-22.04
     steps:
       - name: Checkout code
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/ci.yaml

@@ -182,7 +182,7 @@ jobs:
     runs-on: ubuntu-22.04
     steps:
       - name: Checkout code
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/cleanup-resources.yaml

@@ -16,7 +16,7 @@ jobs:
       name: ci
       deployment: false
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
 

.github/workflows/codeql.yml

@@ -60,7 +60,7 @@ jobs:
         # your codebase is analyzed, see https://docs.github.com/en/code-security/code-scanning/creating-an-advanced-setup-for-code-scanning/codeql-code-scanning-for-compiled-languages
     steps:
     - name: Checkout repository
-      uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
       with:
         persist-credentials: false
 

.github/workflows/darwin-tests.yaml

@@ -27,7 +27,7 @@ jobs:
         echo "$HOME/.local/bin" >> "${GITHUB_PATH}"
 
     - name: Checkout code
-      uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
       with:
         persist-credentials: false
 

.github/workflows/docs-url-alive-check.yaml

@@ -19,7 +19,7 @@ jobs:
       run: |
         echo "GOPATH=${GITHUB_WORKSPACE}" >> "$GITHUB_ENV"
     - name: Checkout code
-      uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
       with:
         fetch-depth: 0
         persist-credentials: false

.github/workflows/editorconfig-checker.yaml

@@ -15,7 +15,7 @@ jobs:
     runs-on: ubuntu-24.04
     steps:
       - name: Checkout the code
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           fetch-depth: 0
           persist-credentials: false

.github/workflows/gatekeeper-skipper.yaml

@@ -42,7 +42,7 @@ jobs:
       skip_test: ${{ steps.skipper.outputs.skip_test }}
       skip_static: ${{ steps.skipper.outputs.skip_static }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/gatekeeper.yaml

@@ -30,7 +30,7 @@ jobs:
       issues: read
       pull-requests: read
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ github.event.pull_request.head.sha }}
           fetch-depth: 0

.github/workflows/nydus-snapshotter-version-in-sync.yaml

@@ -20,7 +20,7 @@ jobs:
     runs-on: ubuntu-22.04
     steps:
     - name: Checkout code
-      uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
       with:
         persist-credentials: false
     - name: Ensure nydus-snapshotter-version is in sync inside our repo

.github/workflows/payload-after-push.yaml

@@ -145,7 +145,7 @@ jobs:
     needs: [publish-kata-deploy-payload-amd64, publish-kata-deploy-payload-arm64, publish-kata-deploy-payload-s390x, publish-kata-deploy-payload-ppc64le]
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
 
@@ -171,7 +171,7 @@ jobs:
       packages: write # needed to push the helm chart to ghcr.io
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
 

.github/workflows/publish-kata-deploy-payload.yaml

@@ -44,7 +44,7 @@ jobs:
       packages: write
     runs-on: ${{ inputs.runner }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/push-oras-tarball-cache.yaml

@@ -21,7 +21,7 @@ jobs:
       packages: write
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           fetch-depth: 0
           persist-credentials: false

.github/workflows/release-amd64.yaml

@@ -50,7 +50,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
       - name: get-kata-tarball

.github/workflows/release-arm64.yaml

@@ -50,7 +50,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
       - name: get-kata-tarball

.github/workflows/release-ppc64le.yaml

@@ -47,7 +47,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
       - name: get-kata-tarball

.github/workflows/release-s390x.yaml

@@ -51,7 +51,7 @@ jobs:
           username: ${{ vars.QUAY_DEPLOYER_USERNAME }}
           password: ${{ secrets.QUAY_DEPLOYER_PASSWORD }}
 
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
       - name: get-kata-tarball

.github/workflows/release.yaml

@@ -12,7 +12,7 @@ jobs:
       contents: write # needed for the `gh release create` command
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           fetch-depth: 0
           persist-credentials: false
@@ -87,7 +87,7 @@ jobs:
       packages: write # needed to push the multi-arch manifest to ghcr.io
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
 
@@ -124,7 +124,7 @@ jobs:
     runs-on: ubuntu-22.04
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
 
@@ -206,7 +206,7 @@ jobs:
       contents: write # needed for the `gh release` commands
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
 
@@ -224,7 +224,7 @@ jobs:
       contents: write # needed for the `gh release` commands
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
 
@@ -242,7 +242,7 @@ jobs:
       contents: write # needed for the `gh release` commands
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
 
@@ -261,7 +261,7 @@ jobs:
       packages: write # needed to push the helm chart to ghcr.io
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
 
@@ -298,7 +298,7 @@ jobs:
       contents: write # needed for the `gh release` commands
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
 

.github/workflows/run-cri-containerd-tests.yaml

@@ -41,7 +41,7 @@ jobs:
       GOPATH: ${{ github.workspace }}
       KATA_HYPERVISOR: ${{ inputs.vmm }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/run-k8s-tests-on-aks.yaml

@@ -73,7 +73,7 @@ jobs:
       GENPOLICY_PULL_METHOD: ${{ matrix.genpolicy-pull-method }}
       RUNS_ON_AKS: "true"
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/run-k8s-tests-on-arm64.yaml

@@ -46,7 +46,7 @@ jobs:
       K8S_TEST_HOST_TYPE: all
       TARGET_ARCH: "aarch64"
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/run-k8s-tests-on-free-runner.yaml

@@ -63,7 +63,7 @@ jobs:
       CONTAINER_ENGINE_VERSION: ${{ matrix.environment.containerd_version }}
       GH_TOKEN: ${{ github.token }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0
@@ -124,4 +124,3 @@ jobs:
         if: always()
         timeout-minutes: 15
         run: bash tests/integration/kubernetes/gha-run.sh cleanup
-

.github/workflows/run-k8s-tests-on-nvidia-gpu.yaml

@@ -53,7 +53,7 @@ jobs:
       USE_EXPERIMENTAL_SNAPSHOTTER_SETUP: ${{ matrix.environment.name == 'nvidia-gpu-snp' && 'true' || 'false' }}
       K8S_TEST_HOST_TYPE: baremetal
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/run-k8s-tests-on-ppc64le.yaml

@@ -45,7 +45,7 @@ jobs:
       KUBERNETES: ${{ matrix.k8s }}
       TARGET_ARCH: "ppc64le"
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/run-k8s-tests-on-zvsi.yaml

@@ -78,7 +78,7 @@ jobs:
       AUTHENTICATED_IMAGE_USER: ${{ vars.AUTHENTICATED_IMAGE_USER }}
       AUTHENTICATED_IMAGE_PASSWORD: ${{ secrets.AUTHENTICATED_IMAGE_PASSWORD }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/run-kata-coco-stability-tests.yaml

@@ -74,7 +74,7 @@ jobs:
       AUTHENTICATED_IMAGE_PASSWORD: ${{ secrets.AUTHENTICATED_IMAGE_PASSWORD }}
       SNAPSHOTTER: ${{ matrix.snapshotter }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/run-kata-coco-tests.yaml

@@ -71,7 +71,7 @@ jobs:
       GH_ITA_KEY: ${{ secrets.ITA_KEY }}
       AUTO_GENERATE_POLICY: "yes"
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0
@@ -169,7 +169,7 @@ jobs:
       CONTAINER_ENGINE_VERSION: "active"
       GH_TOKEN: ${{ github.token }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0
@@ -292,7 +292,7 @@ jobs:
       K8S_TEST_HOST_TYPE: "all"
       GH_TOKEN: ${{ github.token }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0
@@ -408,7 +408,7 @@ jobs:
       AUTO_GENERATE_POLICY: "no"
       GH_TOKEN: ${{ github.token }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/run-kata-deploy-tests-on-aks.yaml

@@ -62,7 +62,7 @@ jobs:
       KATA_HYPERVISOR: ${{ matrix.vmm }}
       KUBERNETES: "vanilla"
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/run-kata-deploy-tests.yaml

@@ -46,7 +46,7 @@ jobs:
       KATA_HYPERVISOR: ${{ matrix.vmm }}
       KUBERNETES: ${{ matrix.k8s }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/run-kata-monitor-tests.yaml

@@ -40,7 +40,7 @@ jobs:
       #CONTAINERD_VERSION: ${{ matrix.containerd_version }}
       KATA_HYPERVISOR: ${{ matrix.vmm }}
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/run-metrics.yaml

@@ -46,7 +46,7 @@ jobs:
       K8S_TEST_HOST_TYPE: "baremetal"
       KUBERNETES: kubeadm
     steps:
-      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           ref: ${{ inputs.commit-hash }}
           fetch-depth: 0

.github/workflows/scorecard.yaml

@@ -27,7 +27,7 @@ jobs:
 
     steps:
       - name: "Checkout code"
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           persist-credentials: false
 

.github/workflows/shellcheck.yaml

@@ -22,7 +22,7 @@ jobs:
     runs-on: ubuntu-24.04
     steps:
       - name: Checkout the code
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           fetch-depth: 0
           persist-credentials: false

.github/workflows/shellcheck_required.yaml

@@ -23,7 +23,7 @@ jobs:
     runs-on: ubuntu-24.04
     steps:
       - name: Checkout the code
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           fetch-depth: 0
           persist-credentials: false

.github/workflows/static-checks.yaml

@@ -28,7 +28,7 @@ jobs:
     runs-on: ubuntu-22.04
     steps:
       - name: Checkout the code
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           fetch-depth: 0
           persist-credentials: false
@@ -71,7 +71,7 @@ jobs:
             component-path: src/dragonball
     steps:
       - name: Checkout the code
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           fetch-depth: 0
           persist-credentials: false
@@ -115,7 +115,7 @@ jobs:
       packages: write # for push to ghcr.io
     steps:
       - name: Checkout code
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           fetch-depth: 0
           persist-credentials: false
@@ -171,7 +171,7 @@ jobs:
       contents: read  # for checkout
     steps:
       - name: Checkout code
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           fetch-depth: 0
           persist-credentials: false

.github/workflows/zizmor.yaml

@@ -15,7 +15,7 @@ jobs:
     runs-on: ubuntu-22.04
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
         with:
           fetch-depth: 0
           persist-credentials: false

ci/openshift-ci/README.md

@@ -37,6 +37,23 @@ oc adm policy add-scc-to-group anyuid system:authenticated system:serviceaccount
 oc label --overwrite ns default pod-security.kubernetes.io/enforce=privileged pod-security.kubernetes.io/warn=baseline pod-security.kubernetes.io/audit=baseline
 ```
 
+The e2e suite uses a combination of built-in (origin) and external tests. External
+tests include Kubernetes upstream conformance tests from the `hyperkube` image.
+
+To enable external tests, export a variable matching your cluster version:
+
+```bash
+export EXTENSIONS_PAYLOAD_OVERRIDE=$(oc get clusterversion version -o jsonpath='{.status.desired.image}')
+# Optional: limit to hyperkube only (k8s conformance tests, avoids downloading all operator extensions)
+export EXTENSION_BINARY_OVERRIDE_INCLUDE_TAGS="hyperkube"
+```
+
+Alternatively, skip external tests entirely (only OpenShift-specific tests from origin):
+
+```bash
+export OPENSHIFT_SKIP_EXTERNAL_TESTS=1
+```
+
 Now you should be ready to run the openshift-tests. Our CI only uses a subset
 of tests, to get the current ``TEST_SKIPS`` see
 [the pipeline config](https://github.com/openshift/release/tree/master/ci-operator/config/kata-containers/kata-containers).

ci/openshift-ci/peer-pods-azure.sh

@@ -39,6 +39,21 @@ git_sparse_clone() {
   git checkout FETCH_HEAD
 }
 
+#######################
+# Install prerequisites
+#######################
+if ! command -v helm &>/dev/null; then
+	echo "Helm not installed, installing in current location..."
+	PATH="${PWD}:${PATH}"
+	curl -fsSL https://raw.githubusercontent.com/helm/helm/main/scripts/get-helm-3 | HELM_INSTALL_DIR='.' bash -s -- --no-sudo
+fi
+if ! command -v yq &>/dev/null; then
+	echo "yq not installed, installing in current location..."
+	PATH="${PWD}:${PATH}"
+	curl -fsSL https://github.com/mikefarah/yq/releases/latest/download/yq_linux_amd64 -o ./yq
+	chmod +x yq
+fi
+
 ###############################
 # Disable security to allow e2e
 ###############################
@@ -83,7 +98,6 @@ AZURE_REGION=$(az group show --resource-group "${AZURE_RESOURCE_GROUP}" --query
 # Create workload identity
 AZURE_WORKLOAD_IDENTITY_NAME="caa-${AZURE_CLIENT_ID}"
 az identity create   --name "${AZURE_WORKLOAD_IDENTITY_NAME}"   --resource-group "${AZURE_RESOURCE_GROUP}"   --location "${AZURE_REGION}"
-USER_ASSIGNED_CLIENT_ID="$(az identity show --resource-group "${AZURE_RESOURCE_GROUP}" --name "${AZURE_WORKLOAD_IDENTITY_NAME}" --query 'clientId' -otsv)"
 
 
 #############################
@@ -184,84 +198,36 @@ echo "CAA_IMAGE=\"${CAA_IMAGE}\""
 echo "CAA_TAG=\"${CAA_TAG}\""
 echo "PP_IMAGE_ID=\"${PP_IMAGE_ID}\""
 
+# Install cert-manager (prerequisit)
+helm install cert-manager oci://quay.io/jetstack/charts/cert-manager --namespace cert-manager --create-namespace --set crds.enabled=true
+
 # Clone and configure caa
-git_sparse_clone "https://github.com/confidential-containers/cloud-api-adaptor.git" "${CAA_GIT_SHA:-main}" "src/cloud-api-adaptor/install/"
+git_sparse_clone "https://github.com/confidential-containers/cloud-api-adaptor.git" "${CAA_GIT_SHA:-main}" "src/cloud-api-adaptor/install/charts/" "src/peerpod-ctrl/chart" "src/webhook/chart"
 echo "CAA_GIT_SHA=\"$(git rev-parse HEAD)\""
-pushd src/cloud-api-adaptor
-cat <<EOF > install/overlays/azure/workload-identity.yaml
-apiVersion: apps/v1
-kind: DaemonSet
-metadata:
-  name: cloud-api-adaptor-daemonset
-  namespace: confidential-containers-system
-spec:
-  template:
-    metadata:
-      labels:
-        azure.workload.identity/use: "true"
----
+pushd src/cloud-api-adaptor/install/charts/peerpods
+# Use the latest kata-deploy
+yq -i '( .dependencies[] | select(.name == "kata-deploy") ) .version = "0.0.0-dev"' Chart.yaml
+helm dependency update .
+# Create secrets
+kubectl apply -f - << EOF
 apiVersion: v1
-kind: ServiceAccount
+kind: Namespace
 metadata:
-  name: cloud-api-adaptor
-  namespace: confidential-containers-system
-  annotations:
-    azure.workload.identity/client-id: "${USER_ASSIGNED_CLIENT_ID}"
+    name: confidential-containers-system
+    labels:
+        app.kubernetes.io/managed-by: Helm
+    annotations:
+        meta.helm.sh/release-name: peerpods
+        meta.helm.sh/release-namespace: confidential-containers-system
 EOF
-PP_INSTANCE_SIZE="Standard_D2as_v5"
-DISABLECVM="true"
-cat <<EOF > install/overlays/azure/kustomization.yaml
-apiVersion: kustomize.config.k8s.io/v1beta1
-kind: Kustomization
-bases:
-- ../../yamls
-images:
-- name: cloud-api-adaptor
-  newName: "${CAA_IMAGE}"
-  newTag: "${CAA_TAG}"
-generatorOptions:
-  disableNameSuffixHash: true
-configMapGenerator:
-- name: peer-pods-cm
-  namespace: confidential-containers-system
-  literals:
-  - CLOUD_PROVIDER="azure"
-  - AZURE_SUBSCRIPTION_ID="${AZURE_SUBSCRIPTION_ID}"
-  - AZURE_REGION="${PP_REGION}"
-  - AZURE_INSTANCE_SIZE="${PP_INSTANCE_SIZE}"
-  - AZURE_RESOURCE_GROUP="${PP_RESOURCE_GROUP}"
-  - AZURE_SUBNET_ID="${PP_SUBNET_ID}"
-  - AZURE_IMAGE_ID="${PP_IMAGE_ID}"
-  - DISABLECVM="${DISABLECVM}"
-  - PEERPODS_LIMIT_PER_NODE="50"
-secretGenerator:
-- name: peer-pods-secret
-  namespace: confidential-containers-system
-  envs:
-  - service-principal.env
-- name: ssh-key-secret
-  namespace: confidential-containers-system
-  files:
-  - id_rsa.pub
-patchesStrategicMerge:
-- workload-identity.yaml
-EOF
-ssh-keygen -t rsa -f install/overlays/azure/id_rsa -N ''
-echo "AZURE_CLIENT_ID=${AZURE_CLIENT_ID}" > install/overlays/azure/service-principal.env
-echo "AZURE_CLIENT_SECRET=${AZURE_CLIENT_SECRET}" >> install/overlays/azure/service-principal.env
-echo "AZURE_TENANT_ID=${AZURE_TENANT_ID}" >> install/overlays/azure/service-principal.env
-
-# Deploy Operator
-git_sparse_clone "https://github.com/confidential-containers/operator" "${OPERATOR_SHA:-main}" "config/"
-echo "OPERATOR_SHA=\"$(git rev-parse HEAD)\""
-oc apply -k "config/release"
-oc apply -k "config/samples/ccruntime/peer-pods"
-popd
-
-# Deploy CAA
-kubectl apply -k "install/overlays/azure"
-popd
-popd
+kubectl create secret generic my-provider-creds \
+       -n confidential-containers-system \
+       --from-literal=AZURE_CLIENT_ID="$AZURE_CLIENT_ID" \
+       --from-literal=AZURE_CLIENT_SECRET="$AZURE_CLIENT_SECRET" \
+       --from-literal=AZURE_TENANT_ID="$AZURE_TENANT_ID"
+helm install peerpods . -f providers/azure.yaml --set secrets.mode=reference --set secrets.existingSecretName=my-provider-creds --set providerConfigs.azure.AZURE_SUBSCRIPTION_ID="${AZURE_SUBSCRIPTION_ID}" --set providerConfigs.azure.AZURE_REGION="${PP_REGION}" --set providerConfigs.azure.AZURE_INSTANCE_SIZE="Standard_D2as_v5" --set providerConfigs.azure.AZURE_RESOURCE_GROUP="${PP_RESOURCE_GROUP}" --set providerConfigs.azure.AZURE_SUBNET_ID="${PP_SUBNET_ID}" --set providerConfigs.azure.AZURE_IMAGE_ID="${PP_IMAGE_ID}" --set providerConfigs.azure.DISABLECVM="true" --set providerConfigs.azure.PEERPODS_LIMIT_PER_NODE="50" --set kata-deploy.snapshotter.setup= --dependency-update -n confidential-containers-system --create-namespace --wait
+popd	# charts
+popd	# git_sparse_clone CAA
 
 # Wait for runtimeclass
 SECONDS=0

src/libs/kata-types/Cargo.toml

@@ -41,6 +41,7 @@ sysctl = "0.7.1"
 tempfile = "3.19.1"
 test-utils = { path = "../test-utils" }
 nix = "0.26.4"
+rstest = "0.18"
 
 [features]
 default = []

src/libs/kata-types/src/config/hypervisor/mod.rs

@@ -1077,6 +1077,81 @@ impl MemoryInfo {
         if self.default_maxmemory == 0 || u64::from(self.default_maxmemory) > host_memory {
             self.default_maxmemory = host_memory as u32;
         }
+
+        // Apply PowerPC64 memory alignment
+        #[cfg(all(target_arch = "powerpc64", target_endian = "little"))]
+        self.adjust_ppc64_memory_alignment()?;
+
+        Ok(())
+    }
+
+    /// Adjusts memory values for PowerPC64 little-endian systems to meet
+    /// QEMU's 256MB block size alignment requirement.
+    ///
+    /// Ensures default_memory is at least 1024MB and both default_memory
+    /// and default_maxmemory are aligned to 256MB boundaries.
+    /// Returns an error if aligned values would be equal.
+    #[cfg(all(target_arch = "powerpc64", target_endian = "little"))]
+    fn adjust_ppc64_memory_alignment(&mut self) -> Result<()> {
+        const PPC64_MEM_BLOCK_SIZE: u64 = 256;
+        const MIN_MEMORY_MB: u64 = 1024;
+
+        fn align_memory(value: u64) -> u64 {
+            (value / PPC64_MEM_BLOCK_SIZE) * PPC64_MEM_BLOCK_SIZE
+        }
+
+        let mut mem_size = u64::from(self.default_memory);
+        let max_mem_size = u64::from(self.default_maxmemory);
+
+        // Ensure minimum memory size
+        if mem_size < MIN_MEMORY_MB {
+            info!(
+                sl!(),
+                "PowerPC: Increasing default_memory from {}MB to minimum {}MB",
+                mem_size,
+                MIN_MEMORY_MB
+            );
+            mem_size = MIN_MEMORY_MB;
+        }
+
+        // Align both values to 256MB boundaries
+        let aligned_mem = align_memory(mem_size);
+        let aligned_max_mem = align_memory(max_mem_size);
+
+        if aligned_mem != mem_size {
+            info!(
+                sl!(),
+                "PowerPC: Aligned default_memory from {}MB to {}MB", mem_size, aligned_mem
+            );
+        }
+
+        if aligned_max_mem != max_mem_size {
+            info!(
+                sl!(),
+                "PowerPC: Aligned default_maxmemory from {}MB to {}MB",
+                max_mem_size,
+                aligned_max_mem
+            );
+        }
+
+        // Check if aligned values are equal
+        if aligned_max_mem != 0 && aligned_max_mem <= aligned_mem {
+            return Err(std::io::Error::other(format!(
+                "PowerPC: default_maxmemory ({}MB) <= default_memory ({}MB) after alignment. \
+                Requires maxmemory > memory. Please increase default_maxmemory.",
+                aligned_max_mem, aligned_mem
+            )));
+        }
+        info!(
+            sl!(),
+            "PowerPC: Memory alignment applied - memory: {}MB, max_memory: {}MB",
+            aligned_mem,
+            aligned_max_mem
+        );
+
+        self.default_memory = aligned_mem as u32;
+        self.default_maxmemory = aligned_max_mem as u32;
+
         Ok(())
     }
 
@@ -1948,4 +2023,72 @@ mod tests {
             );
         }
     }
+
+    #[cfg(all(target_arch = "powerpc64", target_endian = "little"))]
+    use rstest::rstest;
+
+    #[rstest]
+    #[case::memory_below_minimum(512, 2048, 1024, 2048)]
+    #[case::already_aligned(1024, 2048, 1024, 2048)]
+    #[case::unaligned_rounds_down(1100, 2100, 1024, 2048)]
+    #[cfg(all(target_arch = "powerpc64", target_endian = "little"))]
+    fn test_adjust_ppc64_memory_alignment_success(
+        #[case] input_memory: u32,
+        #[case] input_maxmemory: u32,
+        #[case] expected_memory: u32,
+        #[case] expected_maxmemory: u32,
+    ) {
+        let mut mem = MemoryInfo {
+            default_memory: input_memory,
+            default_maxmemory: input_maxmemory,
+            ..Default::default()
+        };
+
+        let result = mem.adjust_ppc64_memory_alignment();
+        assert!(
+            result.is_ok(),
+            "Expected success but got error: {:?}",
+            result.err()
+        );
+        assert_eq!(
+            mem.default_memory, expected_memory,
+            "Memory not aligned correctly"
+        );
+        assert_eq!(
+            mem.default_maxmemory, expected_maxmemory,
+            "Max memory not aligned correctly"
+        );
+    }
+
+    #[rstest]
+    #[case::equal_after_alignment(1024, 1100, "Requires maxmemory > memory")]
+    #[case::maxmemory_less_than_memory(2048, 1500, "Requires maxmemory > memory")]
+    #[cfg(all(target_arch = "powerpc64", target_endian = "little"))]
+    fn test_adjust_ppc64_memory_alignment_errors(
+        #[case] input_memory: u32,
+        #[case] input_maxmemory: u32,
+        #[case] expected_error_msg: &str,
+    ) {
+        let mut mem = MemoryInfo {
+            default_memory: input_memory,
+            default_maxmemory: input_maxmemory,
+            ..Default::default()
+        };
+
+        let result = mem.adjust_ppc64_memory_alignment();
+        assert!(
+            result.is_err(),
+            "Expected error but got success for memory={}, maxmemory={}",
+            input_memory,
+            input_maxmemory
+        );
+
+        let error_msg = result.unwrap_err().to_string();
+        assert!(
+            error_msg.contains(expected_error_msg),
+            "Error message '{}' does not contain expected text '{}'",
+            error_msg,
+            expected_error_msg
+        );
+    }
 }

src/runtime-rs/Makefile

@@ -33,15 +33,11 @@ test:
 	exit 0
 install: install-runtime install-configs
 else ifeq ($(ARCH), powerpc64le)
-default:
-	@echo "PowerPC 64 LE is not currently supported"
-	exit 0
+default: runtime show-header
 test:
-	@echo "PowerPC 64 LE is not currently supported"
-	exit 0
-install:
-	@echo "PowerPC 64 LE is not currently supported"
+	@echo "powerpc64le is not currently supported"
 	exit 0
+install: install-runtime install-configs
 else ifeq ($(ARCH), riscv64gc)
 default: runtime show-header
 test:

src/runtime-rs/arch/powerpc64le-options.mk

@@ -7,9 +7,6 @@
 MACHINETYPE := pseries
 KERNELPARAMS := cgroup_no_v1=all systemd.unified_cgroup_hierarchy=1
 MACHINEACCELERATORS := "cap-cfpc=broken,cap-sbbc=broken,cap-ibs=broken,cap-large-decr=off,cap-ccf-assist=off"
-CPUFEATURES := pmu=off
+CPUFEATURES :=
 
 QEMUCMD := qemu-system-ppc64
-
-# dragonball binary name
-DBCMD := dragonball

src/runtime-rs/crates/hypervisor/src/device/tap.rs

@@ -147,7 +147,7 @@ impl Tap {
 
         // ioctl is safe since we call it with a valid tap fd and check the return
         // value.
-        let ret = unsafe { ioctl_with_mut_ref(&tuntap, TUNSETIFF(), ifr) };
+        let ret = unsafe { ioctl_with_mut_ref(&tuntap, libc::TUNSETIFF as libc::c_ulong, ifr) };
         if ret < 0 {
             return Err(Error::CreateTap(IoError::last_os_error()));
         }

src/runtime-rs/crates/hypervisor/src/qemu/inner.rs

@@ -617,8 +617,10 @@ impl QemuInner {
         todo!()
     }
 
-    pub(crate) fn set_capabilities(&mut self, _flag: CapabilityBits) {
-        todo!()
+    pub(crate) fn set_capabilities(&mut self, flag: CapabilityBits) {
+        let mut caps = Capabilities::default();
+
+        caps.set(flag)
     }
 
     pub(crate) fn set_guest_memory_block_size(&mut self, size: u32) {

src/runtime/pkg/containerd-shim-v2/start.go

@@ -9,9 +9,9 @@ import (
 	"context"
 	"fmt"
 
+	"github.com/containerd/containerd/api/types/task"
 	"github.com/sirupsen/logrus"
 
-	"github.com/containerd/containerd/api/types/task"
 	"github.com/kata-containers/kata-containers/src/runtime/pkg/katautils"
 )
 
@@ -46,6 +46,19 @@ func startContainer(ctx context.Context, s *service, c *container) (retErr error
 		}
 		go watchSandbox(ctx, s)
 
+		// If no network endpoints were discovered during sandbox creation,
+		// schedule an async rescan. This handles runtimes that configure
+		// networking after task creation (e.g. Docker 26+ configures
+		// networking after the Start response, and prestart hooks may
+		// not have run yet on slower architectures).
+		// RescanNetwork is idempotent — it returns immediately if
+		// endpoints already exist.
+		go func() {
+			if err := s.sandbox.RescanNetwork(s.ctx); err != nil {
+				shimLog.WithError(err).Error("async network rescan failed — container may lack networking")
+			}
+		}()
+
 		// We use s.ctx(`ctx` derived from `s.ctx`) to check for cancellation of the
 		// shim context and the context passed to startContainer for tracing.
 		go watchOOMEvents(ctx, s)

src/runtime/pkg/govmm/.github/workflows/main.yml

@@ -1,30 +0,0 @@
-on: ["pull_request"]
-name: Unit tests
-
-permissions:
-  contents: read
-
-jobs:
-  test:
-    name: test
-    strategy:
-      matrix:
-        go-version: [1.15.x, 1.16.x]
-        os: [ubuntu-22.04]
-    runs-on: ${{ matrix.os }}
-    steps:
-      - name: Install Go
-        uses: actions/setup-go@40f1582b2485089dde7abd97c1529aa768e1baff # v5.6.0
-        with:
-          go-version: ${{ matrix.go-version }}
-      - name: Checkout code
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-        with:
-          persist-credentials: false
-      - name: golangci-lint
-        uses: golangci/golangci-lint-action@4696ba8babb6127d732c3c6dde519db15edab9ea # v6.5.1
-        with:
-          version: latest
-          args: -c .golangci.yml -v
-      - name: go test
-        run: go test ./...

src/runtime/pkg/govmm/.gitignore

@@ -1 +0,0 @@
-*~

src/runtime/pkg/govmm/.golangci.yml

@@ -1,35 +0,0 @@
-# Copyright (c) 2021 Intel Corporation
-#
-# SPDX-License-Identifier: Apache-2.0
-
-run:
-  concurrency: 4
-  deadline: 600s
-  skip-dirs:
-    - vendor
-# Ignore auto-generated protobuf code.
-  skip-files:
-    - ".*\\.pb\\.go$"
-
-linters:
-  disable-all: true
-  enable:
-    - deadcode
-    - gocyclo
-    - gofmt
-    - gosimple
-    - govet
-    - ineffassign
-    - misspell
-    - staticcheck
-    - structcheck
-    - typecheck
-    - unconvert
-    - unused
-    - varcheck
-
-linters-settings:
-  gocyclo:
-    min_complexity: 15
-  unused:
-    check-exported: true

src/runtime/pkg/govmm/.travis.yml

@@ -1,26 +0,0 @@
-language: go
-
-go:
-    - "1.10"
-    - "1.11"
-    - tip
-arch:
-    - s390x
-
-go_import_path: github.com/kata-containers/govmm
-
-matrix:
-  allow_failures:
-  - go: tip
-
-before_install:
-  - go get github.com/alecthomas/gometalinter
-  - gometalinter --install
-  - go get github.com/mattn/goveralls
-
-script:
-   - go env
-   - gometalinter --tests --vendor --disable-all --enable=misspell --enable=vet --enable=ineffassign --enable=gofmt --enable=gocyclo --cyclo-over=15 --enable=golint --enable=errcheck --enable=deadcode --enable=staticcheck -enable=gas ./...
-
-after_success:
-   - $GOPATH/bin/goveralls -repotoken $COVERALLS_TOKEN -v -service=travis-ci

src/runtime/pkg/katautils/create.go

@@ -19,6 +19,7 @@ import (
 	vc "github.com/kata-containers/kata-containers/src/runtime/virtcontainers"
 	vf "github.com/kata-containers/kata-containers/src/runtime/virtcontainers/factory"
 	vcAnnotations "github.com/kata-containers/kata-containers/src/runtime/virtcontainers/pkg/annotations"
+	"github.com/kata-containers/kata-containers/src/runtime/virtcontainers/utils"
 	specs "github.com/opencontainers/runtime-spec/specs-go"
 )
 
@@ -140,6 +141,17 @@ func CreateSandbox(ctx context.Context, vci vc.VC, ociSpec specs.Spec, runtimeCo
 		sandboxConfig.Containers[0].RootFs = rootFs
 	}
 
+	// Docker 26+ may set up networking before task creation instead of using
+	// prestart hooks. The netns path is not in the OCI spec but can be
+	// discovered from Docker's libnetwork hook args which contain the sandbox
+	// ID that maps to /var/run/docker/netns/<sandbox_id>.
+	if sandboxConfig.NetworkConfig.NetworkID == "" && !sandboxConfig.NetworkConfig.DisableNewNetwork {
+		if dockerNetns := utils.DockerNetnsPath(&ociSpec); dockerNetns != "" {
+			sandboxConfig.NetworkConfig.NetworkID = dockerNetns
+			kataUtilsLogger.WithField("netns", dockerNetns).Info("discovered Docker network namespace from hook args")
+		}
+	}
+
 	// Important to create the network namespace before the sandbox is
 	// created, because it is not responsible for the creation of the
 	// netns if it does not exist.

src/runtime/pkg/katautils/network_linux.go

@@ -20,7 +20,9 @@ import (
 	"golang.org/x/sys/unix"
 )
 
-const procMountInfoFile = "/proc/self/mountinfo"
+const (
+	procMountInfoFile = "/proc/self/mountinfo"
+)
 
 // EnterNetNS is free from any call to a go routine, and it calls
 // into runtime.LockOSThread(), meaning it won't be executed in a
@@ -29,27 +31,30 @@ func EnterNetNS(networkID string, cb func() error) error {
 	return vc.EnterNetNS(networkID, cb)
 }
 
-// SetupNetworkNamespace create a network namespace
+// SetupNetworkNamespace creates a network namespace if one is not already
+// provided via NetworkID. When NetworkID is empty and networking is not
+// disabled, a new namespace is created as a placeholder; the actual
+// hypervisor namespace will be discovered later by addAllEndpoints after
+// the VM has started.
 func SetupNetworkNamespace(config *vc.NetworkConfig) error {
 	if config.DisableNewNetwork {
 		kataUtilsLogger.Info("DisableNewNetNs is on, shim and hypervisor are running in the host netns")
 		return nil
 	}
 
-	var err error
-	var n ns.NetNS
-
 	if config.NetworkID == "" {
+		var (
+			err error
+			n   ns.NetNS
+		)
+
 		if rootless.IsRootless() {
 			n, err = rootless.NewNS()
-			if err != nil {
-				return err
-			}
 		} else {
 			n, err = testutils.NewNS()
-			if err != nil {
-				return err
-			}
+		}
+		if err != nil {
+			return err
 		}
 
 		config.NetworkID = n.Path()
@@ -71,11 +76,23 @@ func SetupNetworkNamespace(config *vc.NetworkConfig) error {
 }
 
 const (
-	netNsMountType    = "nsfs"
-	mountTypeFieldIdx = 8
-	mountDestIdx      = 4
+	netNsMountType = "nsfs"
+	mountDestIdx   = 4
 )
 
+// mountinfoFsType finds the filesystem type in a parsed mountinfo line.
+// The mountinfo format has optional tagged fields (shared:, master:, etc.)
+// between field 7 and a "-" separator. The fs type is the field immediately
+// after "-". Returns "" if the separator is not found.
+func mountinfoFsType(fields []string) string {
+	for i, f := range fields {
+		if f == "-" && i+1 < len(fields) {
+			return fields[i+1]
+		}
+	}
+	return ""
+}
+
 // getNetNsFromBindMount returns the network namespace for the bind-mounted path
 func getNetNsFromBindMount(nsPath string, procMountFile string) (string, error) {
 	// Resolve all symlinks in the path as the mountinfo file contains
@@ -100,16 +117,15 @@ func getNetNsFromBindMount(nsPath string, procMountFile string) (string, error)
 		// "711 26 0:3 net:[4026532009] /run/docker/netns/default rw shared:535 - nsfs nsfs rw"
 		//
 		// Reference: https://www.kernel.org/doc/Documentation/filesystems/proc.txt
-
-		// We are interested in the first 9 fields of this file,
-		// to check for the correct mount type.
+		// The "-" separator has a variable position due to optional tagged
+		// fields, so we locate the fs type dynamically.
 		fields := strings.Split(text, " ")
 		if len(fields) < 9 {
 			continue
 		}
 
 		// We check here if the mount type is a network namespace mount type, namely "nsfs"
-		if fields[mountTypeFieldIdx] != netNsMountType {
+		if mountinfoFsType(fields) != netNsMountType {
 			continue
 		}
 

src/runtime/pkg/katautils/network_test.go

@@ -149,3 +149,23 @@ func TestSetupNetworkNamespace(t *testing.T) {
 	err = SetupNetworkNamespace(config)
 	assert.NoError(err)
 }
+
+func TestMountinfoFsType(t *testing.T) {
+	assert := assert.New(t)
+
+	// Standard mountinfo line with optional tagged fields
+	fields := []string{"711", "26", "0:3", "net:[4026532009]", "/run/docker/netns/default", "rw", "shared:535", "-", "nsfs", "nsfs", "rw"}
+	assert.Equal("nsfs", mountinfoFsType(fields))
+
+	// Multiple optional tags before separator
+	fields = []string{"711", "26", "0:3", "net:[4026532009]", "/run/docker/netns/default", "rw", "shared:535", "master:1", "-", "nsfs", "nsfs", "rw"}
+	assert.Equal("nsfs", mountinfoFsType(fields))
+
+	// No separator
+	fields = []string{"711", "26", "0:3", "net:[4026532009]", "/run/docker/netns/default", "rw"}
+	assert.Equal("", mountinfoFsType(fields))
+
+	// Separator at end (malformed)
+	fields = []string{"711", "26", "-"}
+	assert.Equal("", mountinfoFsType(fields))
+}

src/runtime/virtcontainers/interfaces.go

@@ -72,6 +72,8 @@ type VCSandbox interface {
 
 	GetOOMEvent(ctx context.Context) (string, error)
 	GetHypervisorPid() (int, error)
+	// RescanNetwork re-scans the network namespace for late-discovered endpoints.
+	RescanNetwork(ctx context.Context) error
 
 	UpdateRuntimeMetrics() error
 	GetAgentMetrics(ctx context.Context) (string, error)

src/runtime/virtcontainers/network_linux.go

@@ -17,9 +17,11 @@ import (
 	"runtime"
 	"sort"
 	"strconv"
+	"strings"
 	"time"
 
 	"github.com/containernetworking/plugins/pkg/ns"
+	"github.com/sirupsen/logrus"
 	"github.com/vishvananda/netlink"
 	"github.com/vishvananda/netns"
 	otelTrace "go.opentelemetry.io/otel/trace"
@@ -45,6 +47,11 @@ type LinuxNetwork struct {
 	interworkingModel NetInterworkingModel
 	netNSCreated      bool
 	danConfigPath     string
+	// placeholderNetNS holds the path to a placeholder network namespace
+	// that we created but later abandoned in favour of the hypervisor's
+	// netns. If best-effort deletion in addAllEndpoints fails, teardown
+	// retries the cleanup via RemoveEndpoints.
+	placeholderNetNS string
 }
 
 // NewNetwork creates a new Linux Network from a NetworkConfig.
@@ -68,11 +75,11 @@ func NewNetwork(configs ...*NetworkConfig) (Network, error) {
 	}
 
 	return &LinuxNetwork{
-		config.NetworkID,
-		[]Endpoint{},
-		config.InterworkingModel,
-		config.NetworkCreated,
-		config.DanConfigPath,
+		netNSPath:         config.NetworkID,
+		eps:               []Endpoint{},
+		interworkingModel: config.InterworkingModel,
+		netNSCreated:      config.NetworkCreated,
+		danConfigPath:     config.DanConfigPath,
 	}, nil
 }
 
@@ -325,28 +332,91 @@ func (n *LinuxNetwork) GetEndpointsNum() (int, error) {
 // Scan the networking namespace through netlink and then:
 // 1. Create the endpoints for the relevant interfaces found there.
 // 2. Attach them to the VM.
+//
+// If no usable interfaces are found and the hypervisor is running in a
+// different network namespace (e.g. Docker 26+ places QEMU in its own
+// pre-configured namespace), switch to the hypervisor's namespace and
+// rescan there. This handles the case where the OCI spec does not
+// communicate the network namespace path.
 func (n *LinuxNetwork) addAllEndpoints(ctx context.Context, s *Sandbox, hotplug bool) error {
-	netnsHandle, err := netns.GetFromPath(n.netNSPath)
+	endpoints, err := n.scanEndpointsInNs(ctx, s, n.netNSPath, hotplug)
 	if err != nil {
 		return err
 	}
+
+	// If the scan found no usable endpoints, check whether the
+	// hypervisor is running in a different namespace and retry there.
+	if len(endpoints) == 0 && s != nil {
+		if hypervisorNs, ok := n.detectHypervisorNetns(s); ok {
+			networkLogger().WithFields(logrus.Fields{
+				"original_netns":   n.netNSPath,
+				"hypervisor_netns": hypervisorNs,
+			}).Debug("no endpoints in original netns, switching to hypervisor netns")
+
+			origPath := n.netNSPath
+			origCreated := n.netNSCreated
+			n.netNSPath = hypervisorNs
+
+			_, err = n.scanEndpointsInNs(ctx, s, n.netNSPath, hotplug)
+			if err != nil {
+				n.netNSPath = origPath
+				n.netNSCreated = origCreated
+				return err
+			}
+
+			// Clean up the placeholder namespace we created — we're now
+			// using the hypervisor's namespace and the placeholder is empty.
+			// Only clear netNSCreated once deletion succeeds; on failure,
+			// stash the path so RemoveEndpoints can retry during teardown.
+			if origCreated {
+				if delErr := deleteNetNS(origPath); delErr != nil {
+					networkLogger().WithField("netns", origPath).WithError(delErr).Warn("failed to delete placeholder netns, will retry during teardown")
+					n.placeholderNetNS = origPath
+				}
+			}
+			// The hypervisor's namespace was not created by us.
+			n.netNSCreated = false
+		}
+	}
+
+	sort.Slice(n.eps, func(i, j int) bool {
+		return n.eps[i].Name() < n.eps[j].Name()
+	})
+
+	networkLogger().WithField("endpoints", n.eps).Info("endpoints found after scan")
+
+	return nil
+}
+
+// scanEndpointsInNs scans a network namespace for usable (non-loopback,
+// configured) interfaces and adds them as endpoints. Returns the list of
+// newly added endpoints.
+func (n *LinuxNetwork) scanEndpointsInNs(ctx context.Context, s *Sandbox, nsPath string, hotplug bool) ([]Endpoint, error) {
+	netnsHandle, err := netns.GetFromPath(nsPath)
+	if err != nil {
+		return nil, err
+	}
 	defer netnsHandle.Close()
 
 	netlinkHandle, err := netlink.NewHandleAt(netnsHandle)
 	if err != nil {
-		return err
+		return nil, err
 	}
 	defer netlinkHandle.Close()
 
 	linkList, err := netlinkHandle.LinkList()
 	if err != nil {
-		return err
+		return nil, err
 	}
 
+	epsBefore := len(n.eps)
+	var added []Endpoint
+
 	for _, link := range linkList {
 		netInfo, err := networkInfoFromLink(netlinkHandle, link)
 		if err != nil {
-			return err
+			// No rollback needed — no endpoints were added in this iteration yet.
+			return nil, err
 		}
 
 		// Ignore unconfigured network interfaces. These are
@@ -368,22 +438,62 @@ func (n *LinuxNetwork) addAllEndpoints(ctx context.Context, s *Sandbox, hotplug
 			continue
 		}
 
-		if err := doNetNS(n.netNSPath, func(_ ns.NetNS) error {
-			_, err = n.addSingleEndpoint(ctx, s, netInfo, hotplug)
-			return err
+		if err := doNetNS(nsPath, func(_ ns.NetNS) error {
+			ep, addErr := n.addSingleEndpoint(ctx, s, netInfo, hotplug)
+			if addErr == nil {
+				added = append(added, ep)
+			}
+			return addErr
 		}); err != nil {
-			return err
+			// Rollback: remove any endpoints added during this scan
+			// so that a failed scan does not leave partial side effects.
+			n.eps = n.eps[:epsBefore]
+			return nil, err
 		}
-
 	}
 
-	sort.Slice(n.eps, func(i, j int) bool {
-		return n.eps[i].Name() < n.eps[j].Name()
-	})
+	return added, nil
+}
 
-	networkLogger().WithField("endpoints", n.eps).Info("endpoints found after scan")
+// detectHypervisorNetns checks whether the hypervisor process is running in a
+// network namespace different from the one we are currently tracking. If so it
+// returns the procfs path to the hypervisor's netns and true.
+func (n *LinuxNetwork) detectHypervisorNetns(s *Sandbox) (string, bool) {
+	pid, err := s.GetHypervisorPid()
+	if err != nil || pid <= 0 {
+		return "", false
+	}
 
-	return nil
+	// Guard against PID recycling: verify the process belongs to this
+	// sandbox by checking its command line for the sandbox ID. QEMU is
+	// started with -name sandbox-<id>, so the ID will appear in cmdline.
+	// /proc/pid/cmdline uses null bytes as argument separators; replace
+	// them so the substring search works on the joined argument string.
+	cmdlineRaw, err := os.ReadFile(fmt.Sprintf("/proc/%d/cmdline", pid))
+	if err != nil {
+		return "", false
+	}
+	cmdline := strings.ReplaceAll(string(cmdlineRaw), "\x00", " ")
+	if !strings.Contains(cmdline, s.id) {
+		return "", false
+	}
+
+	hypervisorNs := fmt.Sprintf("/proc/%d/ns/net", pid)
+
+	// Compare device and inode numbers. Inode numbers are only unique
+	// within a device, so both must match to confirm the same namespace.
+	var currentStat, hvStat unix.Stat_t
+	if err := unix.Stat(n.netNSPath, &currentStat); err != nil {
+		return "", false
+	}
+	if err := unix.Stat(hypervisorNs, &hvStat); err != nil {
+		return "", false
+	}
+
+	if currentStat.Dev != hvStat.Dev || currentStat.Ino != hvStat.Ino {
+		return hypervisorNs, true
+	}
+	return "", false
 }
 
 func convertDanDeviceToNetworkInfo(device *vctypes.DanDevice) (*NetworkInfo, error) {
@@ -571,6 +681,17 @@ func (n *LinuxNetwork) RemoveEndpoints(ctx context.Context, s *Sandbox, endpoint
 		return deleteNetNS(n.netNSPath)
 	}
 
+	// Retry cleanup of a placeholder namespace whose earlier deletion
+	// failed in addAllEndpoints.
+	if n.placeholderNetNS != "" && endpoints == nil {
+		if delErr := deleteNetNS(n.placeholderNetNS); delErr != nil {
+			networkLogger().WithField("netns", n.placeholderNetNS).WithError(delErr).Warn("failed to delete placeholder netns during teardown")
+		} else {
+			networkLogger().WithField("netns", n.placeholderNetNS).Info("placeholder network namespace deleted")
+			n.placeholderNetNS = ""
+		}
+	}
+
 	return nil
 }
 

src/runtime/virtcontainers/network_linux_test.go

@@ -363,11 +363,11 @@ func TestConvertDanDeviceToNetworkInfo(t *testing.T) {
 func TestAddEndpoints_Dan(t *testing.T) {
 
 	network := &LinuxNetwork{
-		"net-123",
-		[]Endpoint{},
-		NetXConnectDefaultModel,
-		true,
-		"testdata/dan-config.json",
+		netNSPath:         "net-123",
+		eps:               []Endpoint{},
+		interworkingModel: NetXConnectDefaultModel,
+		netNSCreated:      true,
+		danConfigPath:     "testdata/dan-config.json",
 	}
 
 	ctx := context.TODO()

src/runtime/virtcontainers/pkg/vcmock/sandbox.go

@@ -255,6 +255,10 @@ func (s *Sandbox) GetHypervisorPid() (int, error) {
 	return 0, nil
 }
 
+func (s *Sandbox) RescanNetwork(ctx context.Context) error {
+	return nil
+}
+
 func (s *Sandbox) GuestVolumeStats(ctx context.Context, path string) ([]byte, error) {
 	return nil, nil
 }

src/runtime/virtcontainers/sandbox.go

@@ -20,6 +20,7 @@ import (
 	"strings"
 	"sync"
 	"syscall"
+	"time"
 
 	v1 "github.com/containerd/cgroups/stats/v1"
 	v2 "github.com/containerd/cgroups/v2/stats"
@@ -330,6 +331,81 @@ func (s *Sandbox) GetHypervisorPid() (int, error) {
 	return pids[0], nil
 }
 
+// RescanNetwork re-scans the network namespace for endpoints if none have
+// been discovered yet. This is idempotent: if endpoints already exist it
+// returns immediately. It enables Docker 26+ support where networking is
+// configured after task creation but before Start.
+//
+// Docker 26+ configures networking (veth pair, IP addresses) between
+// Create and Start. The interfaces may not be present immediately, so
+// this method polls until they appear or a timeout is reached.
+//
+// When new endpoints are found, the guest agent is informed about the
+// interfaces and routes so that networking becomes functional inside the VM.
+func (s *Sandbox) RescanNetwork(ctx context.Context) error {
+	if s.config.NetworkConfig.DisableNewNetwork {
+		return nil
+	}
+	if len(s.network.Endpoints()) > 0 {
+		return nil
+	}
+
+	const maxWait = 5 * time.Second
+	const pollInterval = 50 * time.Millisecond
+	deadline := time.NewTimer(maxWait)
+	defer deadline.Stop()
+	ticker := time.NewTicker(pollInterval)
+	defer ticker.Stop()
+
+	s.Logger().Debug("waiting for network interfaces in namespace")
+
+	for {
+		if _, err := s.network.AddEndpoints(ctx, s, nil, true); err != nil {
+			return err
+		}
+		if len(s.network.Endpoints()) > 0 {
+			return s.configureGuestNetwork(ctx)
+		}
+
+		select {
+		case <-ctx.Done():
+			return ctx.Err()
+		case <-deadline.C:
+			s.Logger().Warn("no network interfaces found after timeout — networking may be configured by prestart hooks")
+			return nil
+		case <-ticker.C:
+		}
+	}
+}
+
+// configureGuestNetwork informs the guest agent about discovered network
+// endpoints so that interfaces and routes become functional inside the VM.
+func (s *Sandbox) configureGuestNetwork(ctx context.Context) error {
+	endpoints := s.network.Endpoints()
+	s.Logger().WithField("endpoints", len(endpoints)).Info("configuring hotplugged network in guest")
+
+	// Note: ARP neighbors (3rd return value) are not propagated here
+	// because the agent interface only exposes per-entry updates. The
+	// full setupNetworks path in kataAgent handles them; this path is
+	// only reached for late-discovered endpoints where neighbor entries
+	// are populated dynamically by the kernel.
+	interfaces, routes, _, err := generateVCNetworkStructures(ctx, endpoints)
+	if err != nil {
+		return fmt.Errorf("generating network structures: %w", err)
+	}
+	for _, ifc := range interfaces {
+		if _, err := s.agent.updateInterface(ctx, ifc); err != nil {
+			return fmt.Errorf("updating interface %s in guest: %w", ifc.Name, err)
+		}
+	}
+	if len(routes) > 0 {
+		if _, err := s.agent.updateRoutes(ctx, routes); err != nil {
+			return fmt.Errorf("updating routes in guest: %w", err)
+		}
+	}
+	return nil
+}
+
 // GetAllContainers returns all containers.
 func (s *Sandbox) GetAllContainers() []VCContainer {
 	ifa := make([]VCContainer, len(s.containers))

src/runtime/virtcontainers/utils/utils.go

@@ -12,6 +12,7 @@ import (
 	"os"
 	"os/exec"
 	"path/filepath"
+	"regexp"
 	"strings"
 	"syscall"
 	"time"
@@ -493,17 +494,38 @@ func RevertBytes(num uint64) uint64 {
 	return 1024*RevertBytes(a) + b
 }
 
+// dockerLibnetworkSetkey is the hook argument that identifies Docker's
+// network configuration hook. The argument following it is the sandbox ID.
+const dockerLibnetworkSetkey = "libnetwork-setkey"
+
+// dockerNetnsPrefixes are the well-known filesystem paths where the Docker
+// daemon bind-mounts container network namespaces.
+var dockerNetnsPrefixes = []string{"/var/run/docker/netns/", "/run/docker/netns/"}
+
+// validSandboxID matches Docker sandbox IDs: exactly 64 lowercase hex characters.
+var validSandboxID = regexp.MustCompile(`^[0-9a-f]{64}$`)
+
 // IsDockerContainer returns if the container is managed by docker
-// This is done by checking the prestart hook for `libnetwork` arguments.
+// This is done by checking the prestart and createRuntime hooks for
+// `libnetwork` arguments. Docker 26+ may use CreateRuntime hooks
+// instead of the deprecated Prestart hooks.
 func IsDockerContainer(spec *specs.Spec) bool {
 	if spec == nil || spec.Hooks == nil {
 		return false
 	}
 
-	for _, hook := range spec.Hooks.Prestart { //nolint:all
-		for _, arg := range hook.Args {
-			if strings.HasPrefix(arg, "libnetwork") {
-				return true
+	// Check both Prestart (Docker < 26) and CreateRuntime (Docker >= 26) hooks.
+	hookSets := [][]specs.Hook{
+		spec.Hooks.Prestart, //nolint:all
+		spec.Hooks.CreateRuntime,
+	}
+
+	for _, hooks := range hookSets {
+		for _, hook := range hooks {
+			for _, arg := range hook.Args {
+				if strings.HasPrefix(arg, "libnetwork") {
+					return true
+				}
 			}
 		}
 	}
@@ -511,6 +533,50 @@ func IsDockerContainer(spec *specs.Spec) bool {
 	return false
 }
 
+// DockerNetnsPath attempts to discover Docker's pre-created network namespace
+// path from OCI spec hooks. Docker's libnetwork-setkey hook contains the
+// sandbox ID as its second argument, which maps to the netns file under
+// /var/run/docker/netns/<sandbox_id>.
+func DockerNetnsPath(spec *specs.Spec) string {
+	if spec == nil || spec.Hooks == nil {
+		return ""
+	}
+
+	// Search both Prestart and CreateRuntime hooks for libnetwork-setkey.
+	hookSets := [][]specs.Hook{
+		spec.Hooks.Prestart, //nolint:all
+		spec.Hooks.CreateRuntime,
+	}
+
+	for _, hooks := range hookSets {
+		for _, hook := range hooks {
+			for i, arg := range hook.Args {
+				if arg == dockerLibnetworkSetkey && i+1 < len(hook.Args) {
+					sandboxID := hook.Args[i+1]
+					// Docker sandbox IDs are exactly 64 lowercase hex
+					// characters. Reject anything else to prevent path
+					// traversal and unexpected input.
+					if !validSandboxID.MatchString(sandboxID) {
+						continue
+					}
+					// Docker stores netns under well-known paths.
+					// Use Lstat to reject symlinks (which could point
+					// outside the Docker netns directory) and non-regular
+					// files such as directories.
+					for _, prefix := range dockerNetnsPrefixes {
+						nsPath := prefix + sandboxID
+						if fi, err := os.Lstat(nsPath); err == nil && fi.Mode().IsRegular() {
+							return nsPath
+						}
+					}
+				}
+			}
+		}
+	}
+
+	return ""
+}
+
 // GetGuestNUMANodes constructs guest NUMA nodes mapping to host NUMA nodes and host CPUs.
 func GetGuestNUMANodes(numaMapping []string) ([]types.GuestNUMANode, error) {
 	// Add guest NUMA node for each specified subsets of host NUMA nodes.

src/runtime/virtcontainers/utils/utils_test.go

@@ -579,24 +579,178 @@ func TestRevertBytes(t *testing.T) {
 	assert.Equal(expectedNum, num)
 }
 
+// TestIsDockerContainer validates hook-detection logic in isolation.
+// End-to-end Docker→containerd→kata integration is covered by
+// external tests (see tests/integration/kubernetes/).
 func TestIsDockerContainer(t *testing.T) {
 	assert := assert.New(t)
 
+	// nil spec
+	assert.False(IsDockerContainer(nil))
+
+	// nil hooks
+	assert.False(IsDockerContainer(&specs.Spec{}))
+
+	// Unrelated prestart hook
 	ociSpec := &specs.Spec{
 		Hooks: &specs.Hooks{
-			Prestart: []specs.Hook{
-				{
-					Args: []string{
-						"haha",
-					},
-				},
+			Prestart: []specs.Hook{ //nolint:all
+				{Args: []string{"haha"}},
 			},
 		},
 	}
 	assert.False(IsDockerContainer(ociSpec))
 
+	// Prestart hook with libnetwork (Docker < 26)
 	ociSpec.Hooks.Prestart = append(ociSpec.Hooks.Prestart, specs.Hook{ //nolint:all
 		Args: []string{"libnetwork-xxx"},
 	})
 	assert.True(IsDockerContainer(ociSpec))
+
+	// CreateRuntime hook with libnetwork (Docker >= 26)
+	ociSpec2 := &specs.Spec{
+		Hooks: &specs.Hooks{
+			CreateRuntime: []specs.Hook{
+				{Args: []string{"/usr/bin/docker-proxy", "libnetwork-setkey", "abc123", "ctrl"}},
+			},
+		},
+	}
+	assert.True(IsDockerContainer(ociSpec2))
+
+	// CreateRuntime hook without libnetwork
+	ociSpec3 := &specs.Spec{
+		Hooks: &specs.Hooks{
+			CreateRuntime: []specs.Hook{
+				{Args: []string{"/some/other/hook"}},
+			},
+		},
+	}
+	assert.False(IsDockerContainer(ociSpec3))
+}
+
+// TestDockerNetnsPath validates netns path discovery from OCI hook args.
+// This does not test the actual namespace opening or endpoint scanning;
+// see integration tests for full-path coverage.
+func TestDockerNetnsPath(t *testing.T) {
+	assert := assert.New(t)
+
+	// Valid 64-char hex sandbox IDs for test cases.
+	validID := strings.Repeat("ab", 32)   // 64 hex chars
+	validID2 := strings.Repeat("cd", 32)  // another 64 hex chars
+	invalidShortID := "abc123"             // too short
+	invalidUpperID := strings.Repeat("AB", 32) // uppercase rejected
+
+	// nil spec
+	assert.Equal("", DockerNetnsPath(nil))
+
+	// nil hooks
+	assert.Equal("", DockerNetnsPath(&specs.Spec{}))
+
+	// Hook without libnetwork-setkey
+	spec := &specs.Spec{
+		Hooks: &specs.Hooks{
+			Prestart: []specs.Hook{ //nolint:all
+				{Args: []string{"/some/binary", "unrelated"}},
+			},
+		},
+	}
+	assert.Equal("", DockerNetnsPath(spec))
+
+	// Prestart hook with libnetwork-setkey but sandbox ID too short (rejected by regex)
+	spec = &specs.Spec{
+		Hooks: &specs.Hooks{
+			Prestart: []specs.Hook{ //nolint:all
+				{Args: []string{"/usr/bin/proxy", "libnetwork-setkey", invalidShortID, "ctrl"}},
+			},
+		},
+	}
+	assert.Equal("", DockerNetnsPath(spec))
+
+	// Prestart hook with libnetwork-setkey but uppercase hex (rejected by regex)
+	spec = &specs.Spec{
+		Hooks: &specs.Hooks{
+			Prestart: []specs.Hook{ //nolint:all
+				{Args: []string{"/usr/bin/proxy", "libnetwork-setkey", invalidUpperID, "ctrl"}},
+			},
+		},
+	}
+	assert.Equal("", DockerNetnsPath(spec))
+
+	// Prestart hook with valid sandbox ID but netns file doesn't exist on disk
+	spec = &specs.Spec{
+		Hooks: &specs.Hooks{
+			Prestart: []specs.Hook{ //nolint:all
+				{Args: []string{"/usr/bin/proxy", "libnetwork-setkey", validID, "ctrl"}},
+			},
+		},
+	}
+	assert.Equal("", DockerNetnsPath(spec))
+
+	// Prestart hook with libnetwork-setkey and existing netns file — success path
+	tmpDir := t.TempDir()
+	fakeNsDir := filepath.Join(tmpDir, "netns")
+	err := os.MkdirAll(fakeNsDir, 0755)
+	assert.NoError(err)
+	fakeNsFile := filepath.Join(fakeNsDir, validID)
+	err = os.WriteFile(fakeNsFile, []byte{}, 0644)
+	assert.NoError(err)
+
+	// Temporarily override dockerNetnsPrefixes so DockerNetnsPath can find
+	// the netns file we created under the temp directory.
+	origPrefixes := dockerNetnsPrefixes
+	dockerNetnsPrefixes = []string{fakeNsDir + "/"}
+	defer func() { dockerNetnsPrefixes = origPrefixes }()
+
+	spec = &specs.Spec{
+		Hooks: &specs.Hooks{
+			Prestart: []specs.Hook{ //nolint:all
+				{Args: []string{"/usr/bin/proxy", "libnetwork-setkey", validID, "ctrl"}},
+			},
+		},
+	}
+	assert.Equal(fakeNsFile, DockerNetnsPath(spec))
+
+	// Sandbox ID that is a directory rather than a regular file — must be rejected
+	dirID := validID2
+	err = os.MkdirAll(filepath.Join(fakeNsDir, dirID), 0755)
+	assert.NoError(err)
+	spec = &specs.Spec{
+		Hooks: &specs.Hooks{
+			Prestart: []specs.Hook{ //nolint:all
+				{Args: []string{"/usr/bin/proxy", "libnetwork-setkey", dirID, "ctrl"}},
+			},
+		},
+	}
+	assert.Equal("", DockerNetnsPath(spec))
+
+	// CreateRuntime hook with valid sandbox ID — file doesn't exist
+	validID3 := strings.Repeat("ef", 32)
+	spec = &specs.Spec{
+		Hooks: &specs.Hooks{
+			CreateRuntime: []specs.Hook{
+				{Args: []string{"/usr/bin/proxy", "libnetwork-setkey", validID3, "ctrl"}},
+			},
+		},
+	}
+	assert.Equal("", DockerNetnsPath(spec))
+
+	// Hook with libnetwork-setkey as last arg (no sandbox ID follows) — no panic
+	spec = &specs.Spec{
+		Hooks: &specs.Hooks{
+			Prestart: []specs.Hook{ //nolint:all
+				{Args: []string{"libnetwork-setkey"}},
+			},
+		},
+	}
+	assert.Equal("", DockerNetnsPath(spec))
+
+	// Empty args slice
+	spec = &specs.Spec{
+		Hooks: &specs.Hooks{
+			Prestart: []specs.Hook{ //nolint:all
+				{Args: []string{}},
+			},
+		},
+	}
+	assert.Equal("", DockerNetnsPath(spec))
 }

tests/containerd-config-v3.tmpl

@@ -1,124 +0,0 @@
-# Copyright (c) K3s contributors
-#
-# SPDX-License-Identifier: Apache-2.0
-#
-
-{{- /* */ -}}
-# File generated by {{ .Program }}. DO NOT EDIT. Use config-v3.toml.tmpl instead.
-version = 3
-imports = ["__CONTAINERD_IMPORTS_PATH__"]
-root = {{ printf "%q" .NodeConfig.Containerd.Root }}
-state = {{ printf "%q" .NodeConfig.Containerd.State }}
-
-[grpc]
- address = {{ deschemify .NodeConfig.Containerd.Address | printf "%q" }}
-
-[plugins.'io.containerd.internal.v1.opt']
- path = {{ printf "%q" .NodeConfig.Containerd.Opt }}
-
-[plugins.'io.containerd.grpc.v1.cri']
- stream_server_address = "127.0.0.1"
- stream_server_port = "10010"
-
-[plugins.'io.containerd.cri.v1.runtime']
- enable_selinux = {{ .NodeConfig.SELinux }}
- enable_unprivileged_ports = {{ .EnableUnprivileged }}
- enable_unprivileged_icmp = {{ .EnableUnprivileged }}
- device_ownership_from_security_context = {{ .NonrootDevices }}
-
-{{ if .DisableCgroup}}
- disable_cgroup = true
-{{ end }}
-
-{{ if .IsRunningInUserNS }}
- disable_apparmor = true
- restrict_oom_score_adj = true
-{{ end }}
-
-{{ with .NodeConfig.AgentConfig.Snapshotter }}
-[plugins.'io.containerd.cri.v1.images']
- snapshotter = "{{ . }}"
- disable_snapshot_annotations = {{ if eq . "stargz" }}false{{else}}true{{end}}
- use_local_image_pull = true
-{{ end }}
-
-{{ with .NodeConfig.AgentConfig.PauseImage }}
-[plugins.'io.containerd.cri.v1.images'.pinned_images]
- sandbox = "{{ . }}"
-{{ end }}
-
-{{- if or .NodeConfig.AgentConfig.CNIBinDir .NodeConfig.AgentConfig.CNIConfDir }}
-[plugins.'io.containerd.cri.v1.runtime'.cni]
- {{ with .NodeConfig.AgentConfig.CNIBinDir }}bin_dirs = [{{ printf "%q" . }}]{{ end }}
- {{ with .NodeConfig.AgentConfig.CNIConfDir }}conf_dir = {{ printf "%q" . }}{{ end }}
-{{ end }}
-
-{{ if or .NodeConfig.Containerd.BlockIOConfig .NodeConfig.Containerd.RDTConfig }}
-[plugins.'io.containerd.service.v1.tasks-service']
- {{ with .NodeConfig.Containerd.BlockIOConfig }}blockio_config_file = {{ printf "%q" . }}{{ end }}
- {{ with .NodeConfig.Containerd.RDTConfig }}rdt_config_file = {{ printf "%q" . }}{{ end }}
-{{ end }}
-
-{{ with .NodeConfig.DefaultRuntime }}
-[plugins.'io.containerd.cri.v1.runtime'.containerd]
- default_runtime_name = "{{ . }}"
-{{ end }}
-
-[plugins.'io.containerd.cri.v1.runtime'.containerd.runtimes.runc]
- runtime_type = "io.containerd.runc.v2"
-
-[plugins.'io.containerd.cri.v1.runtime'.containerd.runtimes.runc.options]
- SystemdCgroup = {{ .SystemdCgroup }}
-
-[plugins.'io.containerd.cri.v1.runtime'.containerd.runtimes.runhcs-wcow-process]
- runtime_type = "io.containerd.runhcs.v1"
-
-{{ range $k, $v := .ExtraRuntimes }}
-[plugins.'io.containerd.cri.v1.runtime'.containerd.runtimes.'{{ $k }}']
- runtime_type = "{{$v.RuntimeType}}"
-{{ with $v.BinaryName}}
-[plugins.'io.containerd.cri.v1.runtime'.containerd.runtimes.'{{ $k }}'.options]
- BinaryName = {{ printf "%q" . }}
- SystemdCgroup = {{ $.SystemdCgroup }}
-{{ end }}
-{{ end }}
-
-[plugins.'io.containerd.cri.v1.images'.registry]
- config_path = {{ printf "%q" .NodeConfig.Containerd.Registry }}
-
-{{ if .PrivateRegistryConfig }}
-{{ range $k, $v := .PrivateRegistryConfig.Configs }}
-{{ with $v.Auth }}
-[plugins.'io.containerd.cri.v1.images'.registry.configs.'{{ $k }}'.auth]
- {{ with .Username }}username = {{ printf "%q" . }}{{ end }}
- {{ with .Password }}password = {{ printf "%q" . }}{{ end }}
- {{ with .Auth }}auth = {{ printf "%q" . }}{{ end }}
- {{ with .IdentityToken }}identitytoken = {{ printf "%q" . }}{{ end }}
-{{ end }}
-{{ end }}
-{{ end }}
-
-{{ if eq .NodeConfig.AgentConfig.Snapshotter "stargz" }}
-{{ with .NodeConfig.AgentConfig.ImageServiceSocket }}
-[plugins.'io.containerd.snapshotter.v1.stargz']
- cri_keychain_image_service_path = {{ printf "%q" . }}
-
-[plugins.'io.containerd.snapshotter.v1.stargz'.cri_keychain]
- enable_keychain = true
-{{ end }}
-
-[plugins.'io.containerd.snapshotter.v1.stargz'.registry]
- config_path = {{ printf "%q" .NodeConfig.Containerd.Registry }}
-
-{{ if .PrivateRegistryConfig }}
-{{ range $k, $v := .PrivateRegistryConfig.Configs }}
-{{ with $v.Auth }}
-[plugins.'io.containerd.snapshotter.v1.stargz'.registry.configs.'{{ $k }}'.auth]
- {{ with .Username }}username = {{ printf "%q" . }}{{ end }}
- {{ with .Password }}password = {{ printf "%q" . }}{{ end }}
- {{ with .Auth }}auth = {{ printf "%q" . }}{{ end }}
- {{ with .IdentityToken }}identitytoken = {{ printf "%q" . }}{{ end }}
-{{ end }}
-{{ end }}
-{{ end }}
-{{ end }}

tests/functional/kata-deploy/kata-deploy-lifecycle.bats

@@ -0,0 +1,213 @@
+#!/usr/bin/env bats
+#
+# Copyright (c) 2026 NVIDIA Corporation
+#
+# SPDX-License-Identifier: Apache-2.0
+#
+# Kata Deploy Lifecycle Tests
+#
+# Validates kata-deploy behavior during DaemonSet restarts and uninstalls:
+#
+# 1. Artifacts present: After install, kata artifacts exist on the host,
+#    RuntimeClasses are created, and the node is labeled.
+#
+# 2. Restart resilience: Running kata pods must survive a kata-deploy
+#    DaemonSet restart without crashing. (Regression test for #12761)
+#
+# 3. Artifact cleanup: After helm uninstall, kata artifacts must be
+#    fully removed from the host and containerd must remain healthy.
+#
+# Required environment variables:
+#   DOCKER_REGISTRY - Container registry for kata-deploy image
+#   DOCKER_REPO     - Repository name for kata-deploy image
+#   DOCKER_TAG      - Image tag to test
+#   KATA_HYPERVISOR - Hypervisor to test (qemu, clh, etc.)
+#   KUBERNETES      - K8s distribution (microk8s, k3s, rke2, etc.)
+
+load "${BATS_TEST_DIRNAME}/../../common.bash"
+repo_root_dir="${BATS_TEST_DIRNAME}/../../../"
+load "${repo_root_dir}/tests/gha-run-k8s-common.sh"
+
+source "${BATS_TEST_DIRNAME}/lib/helm-deploy.bash"
+
+LIFECYCLE_POD_NAME="kata-lifecycle-test"
+
+# Run a command on the host node's filesystem using a short-lived privileged pod.
+# The host root is mounted at /host inside the pod.
+# Usage: run_on_host "test -d /host/opt/kata && echo YES || echo NO"
+run_on_host() {
+	local cmd="$1"
+	local node_name
+	node_name=$(kubectl get nodes --no-headers -o custom-columns=NAME:.metadata.name | head -1)
+	local pod_name="host-exec-${RANDOM}"
+
+	kubectl run "${pod_name}" \
+		--image=quay.io/kata-containers/alpine-bash-curl:latest \
+		--restart=Never --rm -i \
+		--overrides="{
+			\"spec\": {
+				\"nodeName\": \"${node_name}\",
+				\"activeDeadlineSeconds\": 300,
+				\"tolerations\": [{\"operator\": \"Exists\"}],
+				\"containers\": [{
+					\"name\": \"exec\",
+					\"image\": \"quay.io/kata-containers/alpine-bash-curl:latest\",
+					\"imagePullPolicy\": \"IfNotPresent\",
+					\"command\": [\"sh\", \"-c\", \"${cmd}\"],
+					\"securityContext\": {\"privileged\": true},
+					\"volumeMounts\": [{\"name\": \"host\", \"mountPath\": \"/host\", \"readOnly\": true}]
+				}],
+				\"volumes\": [{\"name\": \"host\", \"hostPath\": {\"path\": \"/\"}}]
+			}
+		}"
+}
+
+setup_file() {
+	ensure_helm
+
+	echo "# Image: ${DOCKER_REGISTRY}/${DOCKER_REPO}:${DOCKER_TAG}" >&3
+	echo "# Hypervisor: ${KATA_HYPERVISOR}" >&3
+	echo "# K8s distribution: ${KUBERNETES}" >&3
+	echo "# Deploying kata-deploy..." >&3
+	deploy_kata
+	echo "# kata-deploy deployed successfully" >&3
+}
+
+@test "Kata artifacts are present on host after install" {
+	echo "# Checking kata artifacts on host..." >&3
+
+	run run_on_host "test -d /host/opt/kata && echo PRESENT || echo MISSING"
+	echo "# /opt/kata directory: ${output}" >&3
+	[[ "${output}" == *"PRESENT"* ]]
+
+	run run_on_host "test -f /host/opt/kata/bin/containerd-shim-kata-v2 && echo FOUND || (test -f /host/opt/kata/runtime-rs/bin/containerd-shim-kata-v2 && echo FOUND || echo MISSING)"
+	echo "# containerd-shim-kata-v2: ${output}" >&3
+	[[ "${output}" == *"FOUND"* ]]
+
+	# RuntimeClasses must exist (filter out AKS-managed ones)
+	local rc_count
+	rc_count=$(kubectl get runtimeclasses --no-headers 2>/dev/null | grep -v "kata-mshv-vm-isolation" | grep -c "kata" || true)
+	echo "# Kata RuntimeClasses: ${rc_count}" >&3
+	[[ ${rc_count} -gt 0 ]]
+
+	# Node must have the kata-runtime label
+	local label
+	label=$(kubectl get nodes -o jsonpath='{.items[0].metadata.labels.katacontainers\.io/kata-runtime}')
+	echo "# Node label katacontainers.io/kata-runtime: ${label}" >&3
+	[[ "${label}" == "true" ]]
+}
+
+@test "DaemonSet restart does not crash running kata pods" {
+	# Create a long-running kata pod
+	cat <<EOF | kubectl apply -f -
+apiVersion: v1
+kind: Pod
+metadata:
+  name: ${LIFECYCLE_POD_NAME}
+spec:
+  runtimeClassName: kata-${KATA_HYPERVISOR}
+  restartPolicy: Always
+  nodeSelector:
+    katacontainers.io/kata-runtime: "true"
+  containers:
+    - name: test
+      image: quay.io/kata-containers/alpine-bash-curl:latest
+      imagePullPolicy: IfNotPresent
+      command: ["sleep", "infinity"]
+EOF
+
+	echo "# Waiting for kata pod to be running..." >&3
+	kubectl wait --for=condition=Ready "pod/${LIFECYCLE_POD_NAME}" --timeout=120s
+
+	# Record pod identity before the DaemonSet restart
+	local pod_uid_before
+	pod_uid_before=$(kubectl get pod "${LIFECYCLE_POD_NAME}" -o jsonpath='{.metadata.uid}')
+	local restart_count_before
+	restart_count_before=$(kubectl get pod "${LIFECYCLE_POD_NAME}" -o jsonpath='{.status.containerStatuses[0].restartCount}')
+	echo "# Pod UID before: ${pod_uid_before}, restarts: ${restart_count_before}" >&3
+
+	# Trigger a DaemonSet restart — this simulates what happens when a user
+	# changes a label, updates a config value, or does a rolling update.
+	echo "# Triggering kata-deploy DaemonSet restart..." >&3
+	kubectl -n "${HELM_NAMESPACE}" rollout restart daemonset/kata-deploy
+
+	echo "# Waiting for DaemonSet rollout to complete..." >&3
+	kubectl -n "${HELM_NAMESPACE}" rollout status daemonset/kata-deploy --timeout=300s
+
+	# On k3s/rke2 the new kata-deploy pod restarts the k3s service as
+	# part of install, which causes a brief API server outage. Wait for
+	# the node to become ready before querying pod status.
+	kubectl wait nodes --timeout=120s --all --for condition=Ready=True
+	echo "# Node is ready after DaemonSet rollout" >&3
+
+	# The kata pod must still be Running with the same UID and no extra restarts.
+	# Retry kubectl through any residual API unavailability.
+	local pod_phase=""
+	local retries=0
+	while [[ ${retries} -lt 30 ]]; do
+		pod_phase=$(kubectl get pod "${LIFECYCLE_POD_NAME}" -o jsonpath='{.status.phase}' 2>/dev/null) && break
+		retries=$((retries + 1))
+		sleep 2
+	done
+	echo "# Pod phase after restart: ${pod_phase}" >&3
+	[[ "${pod_phase}" == "Running" ]]
+
+	local pod_uid_after
+	pod_uid_after=$(kubectl get pod "${LIFECYCLE_POD_NAME}" -o jsonpath='{.metadata.uid}')
+	echo "# Pod UID after: ${pod_uid_after}" >&3
+	[[ "${pod_uid_before}" == "${pod_uid_after}" ]]
+
+	local restart_count_after
+	restart_count_after=$(kubectl get pod "${LIFECYCLE_POD_NAME}" -o jsonpath='{.status.containerStatuses[0].restartCount}')
+	echo "# Restart count after: ${restart_count_after}" >&3
+	[[ "${restart_count_before}" == "${restart_count_after}" ]]
+
+	echo "# SUCCESS: Kata pod survived DaemonSet restart without crashing" >&3
+}
+
+@test "Artifacts are fully cleaned up after uninstall" {
+	echo "# Uninstalling kata-deploy..." >&3
+	uninstall_kata
+	echo "# Uninstall complete, verifying cleanup..." >&3
+
+	# Wait for node to recover — containerd restart during cleanup may
+	# cause brief unavailability (especially on k3s/rke2).
+	kubectl wait nodes --timeout=120s --all --for condition=Ready=True
+
+	# RuntimeClasses must be gone (filter out AKS-managed ones)
+	local rc_count
+	rc_count=$(kubectl get runtimeclasses --no-headers 2>/dev/null | grep -v "kata-mshv-vm-isolation" | grep -c "kata" || true)
+	echo "# Kata RuntimeClasses remaining: ${rc_count}" >&3
+	[[ ${rc_count} -eq 0 ]]
+
+	# Node label must be removed
+	local label
+	label=$(kubectl get nodes -o jsonpath='{.items[0].metadata.labels.katacontainers\.io/kata-runtime}' 2>/dev/null || echo "")
+	echo "# Node label after uninstall: '${label}'" >&3
+	[[ -z "${label}" ]]
+
+	# Kata artifacts must be removed from the host filesystem
+	echo "# Checking host filesystem for leftover artifacts..." >&3
+	run run_on_host "test -d /host/opt/kata && echo EXISTS || echo REMOVED"
+	echo "# /opt/kata: ${output}" >&3
+	[[ "${output}" == *"REMOVED"* ]]
+
+	# Containerd must still be healthy and reporting a valid version
+	local container_runtime_version
+	container_runtime_version=$(kubectl get nodes --no-headers -o custom-columns=CONTAINER_RUNTIME:.status.nodeInfo.containerRuntimeVersion)
+	echo "# Container runtime version: ${container_runtime_version}" >&3
+	[[ "${container_runtime_version}" != *"Unknown"* ]]
+
+	echo "# SUCCESS: All kata artifacts cleaned up, containerd healthy" >&3
+}
+
+teardown() {
+	if [[ "${BATS_TEST_NAME}" == *"restart"* ]]; then
+		kubectl delete pod "${LIFECYCLE_POD_NAME}" --ignore-not-found=true --wait=false 2>/dev/null || true
+	fi
+}
+
+teardown_file() {
+	kubectl delete pod "${LIFECYCLE_POD_NAME}" --ignore-not-found=true --wait=false 2>/dev/null || true
+	uninstall_kata 2>/dev/null || true
+}

tests/functional/kata-deploy/run-kata-deploy-tests.sh

@@ -20,6 +20,7 @@ else
 	KATA_DEPLOY_TEST_UNION=( \
 		"kata-deploy.bats" \
 		"kata-deploy-custom-runtimes.bats" \
+		"kata-deploy-lifecycle.bats" \
 	)
 fi
 

tests/gha-run-k8s-common.sh

@@ -296,36 +296,6 @@ function deploy_k0s() {
 	sudo chown "${USER}":"${USER}" ~/.kube/config
 }
 
-# If the rendered containerd config (v3) does not import the drop-in dir, write
-# the full V3 template (from tests/containerd-config-v3.tmpl) with the given
-# import path and restart the service.
-# Args: containerd_dir (e.g. /var/lib/rancher/k3s/agent/etc/containerd), service_name (e.g. k3s or rke2-server).
-function _setup_containerd_v3_template_if_needed() {
-	local containerd_dir="$1"
-	local service_name="$2"
-	local template_file="${tests_dir}/containerd-config-v3.tmpl"
-	local rendered_v3="${containerd_dir}/config-v3.toml"
-	local imports_path="${containerd_dir}/config-v3.toml.d/*.toml"
-	if sudo test -f "${rendered_v3}" && sudo grep -q 'config-v3\.toml\.d' "${rendered_v3}" 2>/dev/null; then
-		return 0
-	fi
-	if [[ ! -f "${template_file}" ]]; then
-		echo "Template not found: ${template_file}" >&2
-		return 1
-	fi
-	sudo mkdir -p "${containerd_dir}/config-v3.toml.d"
-	sed "s|__CONTAINERD_IMPORTS_PATH__|${imports_path}|g" "${template_file}" | sudo tee "${containerd_dir}/config-v3.toml.tmpl" > /dev/null
-	sudo systemctl restart "${service_name}"
-}
-
-function setup_k3s_containerd_v3_template_if_needed() {
-	_setup_containerd_v3_template_if_needed "/var/lib/rancher/k3s/agent/etc/containerd" "k3s"
-}
-
-function setup_rke2_containerd_v3_template_if_needed() {
-	_setup_containerd_v3_template_if_needed "/var/lib/rancher/rke2/agent/etc/containerd" "rke2-server"
-}
-
 function deploy_k3s() {
 	# Set CRI runtime-request-timeout to 600s (same as kubeadm) for CoCo and long-running create requests.
 	curl -sfL https://get.k3s.io | sh -s - --write-kubeconfig-mode 644 --kubelet-arg runtime-request-timeout=600s
@@ -333,9 +303,6 @@ function deploy_k3s() {
 	# This is an arbitrary value that came up from local tests
 	sleep 120s
 
-	# If rendered config does not import the drop-in dir, write full V3 template so kata-deploy can use it.
-	setup_k3s_containerd_v3_template_if_needed
-
 	# Download the kubectl binary into /usr/bin and remove /usr/local/bin/kubectl
 	#
 	# We need to do this to avoid hitting issues like:
@@ -405,9 +372,6 @@ function deploy_rke2() {
 	# This is an arbitrary value that came up from local tests
 	sleep 120s
 
-	# If rendered config does not import the drop-in dir, write full V3 template so kata-deploy can use it.
-	setup_rke2_containerd_v3_template_if_needed
-
 	# Link the kubectl binary into /usr/bin
 	sudo ln -sf /var/lib/rancher/rke2/bin/kubectl /usr/local/bin/kubectl
 

tests/integration/docker/gha-run.sh

@@ -0,0 +1,45 @@
+#!/bin/bash
+#
+# Copyright (c) 2023 Intel Corporation
+#
+# SPDX-License-Identifier: Apache-2.0
+#
+
+set -o errexit
+set -o nounset
+set -o pipefail
+
+kata_tarball_dir="${2:-kata-artifacts}"
+docker_dir="$(dirname "$(readlink -f "$0")")"
+source "${docker_dir}/../../common.bash"
+image="${image:-instrumentisto/nmap:latest}"
+
+function install_dependencies() {
+	info "Installing the dependencies needed for running the docker smoke test"
+
+	sudo -E docker pull "${image}"
+}
+
+function run() {
+	info "Running docker smoke test tests using ${KATA_HYPERVISOR} hypervisor"
+
+	enabling_hypervisor
+
+	info "Running docker with runc"
+	sudo docker run --rm --entrypoint nping "${image}" --tcp-connect -c 2 -p 80 www.github.com
+
+	info "Running docker with Kata Containers (${KATA_HYPERVISOR})"
+	sudo docker run --rm --runtime io.containerd.kata-${KATA_HYPERVISOR}.v2 --entrypoint nping "${image}" --tcp-connect -c 2 -p 80 www.github.com
+}
+
+function main() {
+	action="${1:-}"
+	case "${action}" in
+		install-dependencies) install_dependencies ;;
+		install-kata) install_kata ;;
+		run) run ;;
+		*) >&2 die "Invalid argument" ;;
+	esac
+}
+
+main "$@"

tests/integration/kubernetes/confidential_common.sh

@@ -116,12 +116,16 @@ function is_confidential_gpu_hardware() {
 	return 1
 }
 
+# create_loop_device creates a loop device backed by a file.
+# $1: loop file path (default: /tmp/trusted-image-storage.img)
+# $2: size in MiB, i.e. dd bs=1M count=... (default: 2500, ~2.4Gi)
 function create_loop_device(){
 	local loop_file="${1:-/tmp/trusted-image-storage.img}"
+	local size_mb="${2:-2500}"
 	local node="$(get_one_kata_node)"
 	cleanup_loop_device "$loop_file"
 
-	exec_host "$node" "dd if=/dev/zero of=$loop_file bs=1M count=2500"
+	exec_host "$node" "dd if=/dev/zero of=$loop_file bs=1M count=$size_mb"
 	exec_host "$node" "losetup -fP $loop_file >/dev/null 2>&1"
 	local device=$(exec_host "$node" losetup -j $loop_file | awk -F'[: ]' '{print $1}')
 

tests/integration/kubernetes/k8s-guest-pull-image.bats

@@ -97,7 +97,10 @@ setup() {
 
     storage_config=$(mktemp "${BATS_FILE_TMPDIR}/$(basename "${storage_config_template}").XXXXXX.yaml")
     local_device=$(create_loop_device)
-    LOCAL_DEVICE="$local_device" NODE_NAME="$node" envsubst < "$storage_config_template" > "$storage_config"
+    PV_NAME=trusted-block-pv PVC_NAME=trusted-pvc \
+        PV_STORAGE_CAPACITY=10Gi PVC_STORAGE_REQUEST=1Gi \
+        LOCAL_DEVICE="$local_device" NODE_NAME="$node" \
+        envsubst < "$storage_config_template" > "$storage_config"
 
     # For debug sake
     echo "Trusted storage $storage_config file:"
@@ -142,7 +145,10 @@ setup() {
 @test "Test we cannot pull a large image that pull time exceeds createcontainer timeout inside the guest" {
     storage_config=$(mktemp "${BATS_FILE_TMPDIR}/$(basename "${storage_config_template}").XXXXXX.yaml")
     local_device=$(create_loop_device)
-    LOCAL_DEVICE="$local_device" NODE_NAME="$node" envsubst < "$storage_config_template" > "$storage_config"
+    PV_NAME=trusted-block-pv PVC_NAME=trusted-pvc \
+        PV_STORAGE_CAPACITY=10Gi PVC_STORAGE_REQUEST=1Gi \
+        LOCAL_DEVICE="$local_device" NODE_NAME="$node" \
+        envsubst < "$storage_config_template" > "$storage_config"
 
     # For debug sake
     echo "Trusted storage $storage_config file:"
@@ -193,7 +199,10 @@ setup() {
     fi
     storage_config=$(mktemp "${BATS_FILE_TMPDIR}/$(basename "${storage_config_template}").XXXXXX.yaml")
     local_device=$(create_loop_device)
-    LOCAL_DEVICE="$local_device" NODE_NAME="$node" envsubst < "$storage_config_template" > "$storage_config"
+    PV_NAME=trusted-block-pv PVC_NAME=trusted-pvc \
+        PV_STORAGE_CAPACITY=10Gi PVC_STORAGE_REQUEST=1Gi \
+        LOCAL_DEVICE="$local_device" NODE_NAME="$node" \
+        envsubst < "$storage_config_template" > "$storage_config"
 
     # For debug sake
     echo "Trusted storage $storage_config file:"

tests/integration/kubernetes/k8s-nvidia-nim-service.bats

@@ -0,0 +1,219 @@
+#!/usr/bin/env bats
+#
+# Copyright (c) 2026 NVIDIA Corporation
+#
+# SPDX-License-Identifier: Apache-2.0
+#
+# This file is modeled after k8s-nvidia-nim.bats which contains helpful in-line documentation.
+
+load "${BATS_TEST_DIRNAME}/lib.sh"
+load "${BATS_TEST_DIRNAME}/confidential_common.sh"
+
+export KATA_HYPERVISOR="${KATA_HYPERVISOR:-qemu-nvidia-gpu}"
+
+TEE=false
+if is_confidential_gpu_hardware; then
+    TEE=true
+fi
+export TEE
+
+NIM_SERVICE_NAME="meta-llama-3-2-1b-instruct"
+[[ "${TEE}" = "true" ]] && NIM_SERVICE_NAME="meta-llama-3-2-1b-instruct-tee"
+export NIM_SERVICE_NAME
+
+POD_READY_TIMEOUT_LLAMA_3_2_1B_PREDEFINED=600s
+[[ "${TEE}" = "true" ]] && POD_READY_TIMEOUT_LLAMA_3_2_1B_PREDEFINED=1200s
+export POD_READY_TIMEOUT_LLAMA_3_2_1B=${POD_READY_TIMEOUT_LLAMA_3_2_1B:-${POD_READY_TIMEOUT_LLAMA_3_2_1B_PREDEFINED}}
+
+export LOCAL_NIM_CACHE_LLAMA_3_2_1B="${LOCAL_NIM_CACHE_LLAMA_3_2_1B:-${LOCAL_NIM_CACHE:-/opt/nim/.cache}-llama-3-2-1b}"
+
+DOCKER_CONFIG_JSON=$(
+    echo -n "{\"auths\":{\"nvcr.io\":{\"username\":\"\$oauthtoken\",\"password\":\"${NGC_API_KEY}\",\"auth\":\"$(echo -n "\$oauthtoken:${NGC_API_KEY}" | base64 -w0)\"}}}" |
+        base64 -w0
+)
+export DOCKER_CONFIG_JSON
+
+KBS_AUTH_CONFIG_JSON=$(
+    echo -n "{\"auths\":{\"nvcr.io\":{\"auth\":\"$(echo -n "\$oauthtoken:${NGC_API_KEY}" | base64 -w0)\"}}}" |
+        base64 -w0
+)
+export KBS_AUTH_CONFIG_JSON
+
+NGC_API_KEY_BASE64=$(
+    echo -n "${NGC_API_KEY}" | base64 -w0
+)
+export NGC_API_KEY_BASE64
+
+# Points to kbs:///default/ngc-api-key/instruct and thus re-uses the secret from k8s-nvidia-nim.bats.
+NGC_API_KEY_SEALED_SECRET_LLAMA_3_2_1B="${SEALED_SECRET_PRECREATED_NIM_INSTRUCT}"
+export NGC_API_KEY_SEALED_SECRET_LLAMA_3_2_1B
+
+NGC_API_KEY_SEALED_SECRET_LLAMA_3_2_1B_BASE64=$(echo -n "${NGC_API_KEY_SEALED_SECRET_LLAMA_3_2_1B}" | base64 -w0)
+export NGC_API_KEY_SEALED_SECRET_LLAMA_3_2_1B_BASE64
+
+# NIM Operator (k8s-nim-operator) install/uninstall for NIMService CRD.
+NIM_OPERATOR_NAMESPACE="${NIM_OPERATOR_NAMESPACE:-nim-operator}"
+NIM_OPERATOR_RELEASE_NAME="nim-operator"
+
+install_nim_operator() {
+	command -v helm &>/dev/null || die "helm is required but not installed"
+	echo "Installing NVIDIA NIM Operator (latest chart)"
+	helm repo add nvidia https://helm.ngc.nvidia.com/nvidia
+	helm repo update
+
+	kubectl create namespace "${NIM_OPERATOR_NAMESPACE}" --dry-run=client -o yaml | kubectl apply -f -
+
+	helm upgrade --install "${NIM_OPERATOR_RELEASE_NAME}" nvidia/k8s-nim-operator \
+		-n "${NIM_OPERATOR_NAMESPACE}" \
+		--wait
+
+	local deploy_name
+	deploy_name=$(kubectl get deployment -n "${NIM_OPERATOR_NAMESPACE}" -o jsonpath='{.items[0].metadata.name}' 2>/dev/null || true)
+	if [[ -n "${deploy_name}" ]]; then
+		kubectl wait --for=condition=available --timeout=300s "deployment/${deploy_name}" -n "${NIM_OPERATOR_NAMESPACE}"
+	fi
+	echo "NIM Operator install complete."
+}
+
+uninstall_nim_operator() {
+	echo "Uninstalling NVIDIA NIM Operator (release: ${NIM_OPERATOR_RELEASE_NAME}, namespace: ${NIM_OPERATOR_NAMESPACE})"
+	if helm status "${NIM_OPERATOR_RELEASE_NAME}" -n "${NIM_OPERATOR_NAMESPACE}" &>/dev/null; then
+		helm uninstall "${NIM_OPERATOR_RELEASE_NAME}" -n "${NIM_OPERATOR_NAMESPACE}" || true
+		kubectl delete namespace "${NIM_OPERATOR_NAMESPACE}" --ignore-not-found=true --timeout=60s || true
+		echo "NIM Operator uninstall complete."
+	else
+		echo "NIM Operator release not found, nothing to uninstall."
+	fi
+}
+
+setup_kbs_credentials() {
+    CC_KBS_ADDR=$(kbs_k8s_svc_http_addr)
+    export CC_KBS_ADDR
+
+    kubectl delete secret ngc-secret-llama-3-2-1b --ignore-not-found
+    kubectl create secret docker-registry ngc-secret-llama-3-2-1b --docker-server="nvcr.io" --docker-username="\$oauthtoken" --docker-password="${NGC_API_KEY}"
+
+    kbs_set_gpu0_resource_policy
+    kbs_set_resource_base64 "default" "credentials" "nvcr" "${KBS_AUTH_CONFIG_JSON}"
+    kbs_set_resource "default" "ngc-api-key" "instruct" "${NGC_API_KEY}"
+}
+
+# CDH initdata for guest-pull: KBS URL, registry credentials URI, and allow-all policy.
+# NIMService is not supported by genpolicy; add_allow_all_policy_to_yaml only supports Pod/Deployment.
+# Build initdata with policy inline so TEE pods get both CDH config and policy.
+create_nim_initdata_file_llama_3_2_1b() {
+    local output_file="$1"
+    local cc_kbs_address
+    cc_kbs_address=$(kbs_k8s_svc_http_addr)
+    local allow_all_rego="${BATS_TEST_DIRNAME}/../../../src/kata-opa/allow-all.rego"
+
+    cat > "${output_file}" << EOF
+version = "0.1.0"
+algorithm = "sha256"
+
+[data]
+"aa.toml" = '''
+[token_configs]
+[token_configs.kbs]
+url = "${cc_kbs_address}"
+'''
+
+"cdh.toml" = '''
+[kbc]
+name = "cc_kbc"
+url = "${cc_kbs_address}"
+
+[image]
+authenticated_registry_credentials_uri = "kbs:///default/credentials/nvcr"
+'''
+
+"policy.rego" = '''
+$(cat "${allow_all_rego}")
+'''
+EOF
+}
+
+setup() {
+    setup_common || die "setup_common failed"
+    install_nim_operator || die "NIM Operator install failed"
+
+    dpkg -s jq >/dev/null 2>&1 || sudo apt -y install jq
+
+    # Same pattern as k8s-nvidia-nim.bats: choose manifest by TEE; each YAML has literal secret names.
+    local tee_suffix=""
+    [[ "${TEE}" = "true" ]] && tee_suffix="-tee"
+    export NIM_YAML_IN="${pod_config_dir}/nvidia-nim-llama-3-2-1b-instruct-service${tee_suffix}.yaml.in"
+    export NIM_YAML="${pod_config_dir}/nvidia-nim-llama-3-2-1b-instruct-service${tee_suffix}.yaml"
+
+    if [[ "${TEE}" = "true" ]]; then
+        setup_kbs_credentials
+        setup_sealed_secret_signing_public_key
+        initdata_file="${BATS_SUITE_TMPDIR}/nim-initdata-llama-3-2-1b.toml"
+        create_nim_initdata_file_llama_3_2_1b "${initdata_file}"
+        NIM_INITDATA_BASE64=$(gzip -c "${initdata_file}" | base64 -w0)
+        export NIM_INITDATA_BASE64
+    fi
+
+    envsubst < "${NIM_YAML_IN}" > "${NIM_YAML}"
+}
+
+@test "NIMService llama-3.2-1b-instruct serves /v1/models" {
+    echo "NIMService test: Applying NIM YAML"
+    kubectl apply -f "${NIM_YAML}"
+    echo "NIMService test: Waiting for deployment to exist (operator creates it from NIMService)"
+    local wait_exist_timeout=30
+    local elapsed=0
+    while ! kubectl get deployment "${NIM_SERVICE_NAME}" &>/dev/null; do
+        if [[ ${elapsed} -ge ${wait_exist_timeout} ]]; then
+            echo "Deployment ${NIM_SERVICE_NAME} did not appear within ${wait_exist_timeout}s" >&2
+            kubectl get deployment "${NIM_SERVICE_NAME}" 2>&1 || true
+            false
+        fi
+        sleep 5
+        elapsed=$((elapsed + 5))
+    done
+    local pod_name
+    pod_name=$(kubectl get pods --no-headers -o custom-columns=":metadata.name" | head -1)
+    echo "NIMService test: POD_NAME=${pod_name} (waiting for pod ready, timeout ${POD_READY_TIMEOUT_LLAMA_3_2_1B})"
+    [[ -n "${pod_name}" ]]
+    kubectl wait --for=condition=ready --timeout="${POD_READY_TIMEOUT_LLAMA_3_2_1B}" "pod/${pod_name}"
+    local pod_ip
+    pod_ip=$(kubectl get pod "${pod_name}" -o jsonpath='{.status.podIP}')
+    echo "NIMService test: POD_IP=${pod_ip}"
+    [[ -n "${pod_ip}" ]]
+
+    echo "NIMService test: Curling http://${pod_ip}:8000/v1/models"
+    run curl -sS --connect-timeout 10 "http://${pod_ip}:8000/v1/models"
+    echo "NIMService test: /v1/models response: ${output}"
+    [[ "${status}" -eq 0 ]]
+    [[ "$(echo "${output}" | jq -r '.object')" == "list" ]]
+    [[ "$(echo "${output}" | jq -r '.data[0].id')" == "meta/llama-3.2-1b-instruct" ]]
+    [[ "$(echo "${output}" | jq -r '.data[0].object')" == "model" ]]
+
+    echo "NIMService test: Curling http://${pod_ip}:8000/v1/chat/completions"
+    run curl -sS --connect-timeout 30 "http://${pod_ip}:8000/v1/chat/completions" \
+        -H "Content-Type: application/json" \
+        -d '{"model":"meta/llama-3.2-1b-instruct","messages":[{"role":"user","content":"ping"}],"max_tokens":8}'
+    echo "NIMService test: /v1/chat/completions response: ${output}"
+    [[ "${status}" -eq 0 ]]
+    [[ "$(echo "${output}" | jq -r '.object')" == "chat.completion" ]]
+    [[ "$(echo "${output}" | jq -r '.model')" == "meta/llama-3.2-1b-instruct" ]]
+    [[ "$(echo "${output}" | jq -r '.choices[0].message | has("content") or has("reasoning_content")')" == "true" ]]
+}
+
+teardown() {
+    if kubectl get nimservice "${NIM_SERVICE_NAME}" &>/dev/null; then
+        POD_NAME=$(kubectl get pods --no-headers -o custom-columns=":metadata.name" | head -1)
+        if [[ -n "${POD_NAME}" ]]; then
+            echo "=== NIMService pod logs ==="
+            kubectl logs "${POD_NAME}" || true
+            kubectl describe pod "${POD_NAME}" || true
+        fi
+        kubectl describe nimservice "${NIM_SERVICE_NAME}" || true
+    fi
+
+    [ -f "${NIM_YAML}" ] && kubectl delete -f "${NIM_YAML}" --ignore-not-found=true
+
+    uninstall_nim_operator || true
+    print_node_journal_since_test_start "${node}" "${node_start_time:-}" "${BATS_TEST_COMPLETED:-}"
+}

tests/integration/kubernetes/k8s-nvidia-nim.bats

@@ -85,6 +85,8 @@ setup_langchain_flow() {
 # generated policy.rego to it and set it as the cc_init_data annotation.
 # We must overwrite the default empty file AFTER create_tmp_policy_settings_dir()
 # copies it to the temp directory.
+# As we use multiple vCPUs we set `max_concurrent_layer_downloads_per_image = 1`,
+# see: https://github.com/kata-containers/kata-containers/issues/12721
 create_nim_initdata_file() {
     local output_file="$1"
     local cc_kbs_address
@@ -107,6 +109,7 @@ name = "cc_kbc"
 url = "${cc_kbs_address}"
 
 [image]
+max_concurrent_layer_downloads_per_image = 1
 authenticated_registry_credentials_uri = "kbs:///default/credentials/nvcr"
 '''
 EOF
@@ -189,12 +192,35 @@ setup_file() {
         # This must happen AFTER create_tmp_policy_settings_dir() copies the empty
         # file and BEFORE auto_generate_policy() runs.
         create_nim_initdata_file "${policy_settings_dir}/default-initdata.toml"
+
+        # Container image layer storage: one block device and PV/PVC per pod.
+        storage_config_template="${pod_config_dir}/confidential/trusted-storage.yaml.in"
+
+        instruct_storage_mib=57344
+        local_device_instruct=$(create_loop_device /tmp/trusted-image-storage-instruct.img "$instruct_storage_mib")
+        storage_config_instruct=$(mktemp "${BATS_FILE_TMPDIR}/$(basename "${storage_config_template}").instruct.XXX")
+        PV_NAME=trusted-block-pv-instruct PVC_NAME=trusted-pvc-instruct \
+            PV_STORAGE_CAPACITY="${instruct_storage_mib}Mi" PVC_STORAGE_REQUEST="${instruct_storage_mib}Mi" \
+            LOCAL_DEVICE="$local_device_instruct" NODE_NAME="$node" \
+            envsubst < "$storage_config_template" > "$storage_config_instruct"
+        retry_kubectl_apply "$storage_config_instruct"
+
+        if [ "${SKIP_MULTI_GPU_TESTS}" != "true" ]; then
+            embedqa_storage_mib=8192
+            local_device_embedqa=$(create_loop_device /tmp/trusted-image-storage-embedqa.img "$embedqa_storage_mib")
+            storage_config_embedqa=$(mktemp "${BATS_FILE_TMPDIR}/$(basename "${storage_config_template}").embedqa.XXX")
+            PV_NAME=trusted-block-pv-embedqa PVC_NAME=trusted-pvc-embedqa \
+                PV_STORAGE_CAPACITY="${embedqa_storage_mib}Mi" PVC_STORAGE_REQUEST="${embedqa_storage_mib}Mi" \
+                LOCAL_DEVICE="$local_device_embedqa" NODE_NAME="$node" \
+                envsubst < "$storage_config_template" > "$storage_config_embedqa"
+            retry_kubectl_apply "$storage_config_embedqa"
+        fi
     fi
 
     create_inference_pod
 
     if [ "${SKIP_MULTI_GPU_TESTS}" != "true" ]; then
-         create_embedqa_pod
+        create_embedqa_pod
     fi
 }
 
@@ -459,5 +485,13 @@ teardown_file() {
         [ -f "${POD_EMBEDQA_YAML}" ] && kubectl delete -f "${POD_EMBEDQA_YAML}" --ignore-not-found=true
     fi
 
+    if [[ "${TEE}" = "true" ]]; then
+        kubectl delete --ignore-not-found pvc trusted-pvc-instruct trusted-pvc-embedqa
+        kubectl delete --ignore-not-found pv trusted-block-pv-instruct trusted-block-pv-embedqa
+        kubectl delete --ignore-not-found storageclass local-storage
+        cleanup_loop_device /tmp/trusted-image-storage-instruct.img || true
+        cleanup_loop_device /tmp/trusted-image-storage-embedqa.img || true
+    fi
+
     print_node_journal_since_test_start "${node}" "${node_start_time:-}" "${BATS_TEST_COMPLETED:-}" >&3
 }

tests/integration/kubernetes/run_kubernetes_nv_tests.sh

@@ -89,7 +89,8 @@ if [[ -n "${K8S_TEST_NV:-}" ]]; then
 else
 	K8S_TEST_NV=("k8s-confidential-attestation.bats" \
 		"k8s-nvidia-cuda.bats" \
-		"k8s-nvidia-nim.bats")
+		"k8s-nvidia-nim.bats" \
+		"k8s-nvidia-nim-service.bats")
 fi
 
 SUPPORTED_HYPERVISORS=("qemu-nvidia-gpu" "qemu-nvidia-gpu-snp" "qemu-nvidia-gpu-tdx")

tests/integration/kubernetes/runtimeclass_workloads/confidential/trusted-storage.yaml.in

@@ -14,10 +14,10 @@ volumeBindingMode: WaitForFirstConsumer
 apiVersion: v1
 kind: PersistentVolume
 metadata:
-  name: trusted-block-pv
+  name: $PV_NAME
 spec:
   capacity:
-    storage: 10Gi
+    storage: $PV_STORAGE_CAPACITY
   volumeMode: Block
   accessModes:
     - ReadWriteOnce
@@ -37,12 +37,12 @@ spec:
 apiVersion: v1
 kind: PersistentVolumeClaim
 metadata:
-  name: trusted-pvc
+  name: $PVC_NAME
 spec:
   accessModes:
     - ReadWriteOnce
   resources:
     requests:
-      storage: 1Gi
+      storage: $PVC_STORAGE_REQUEST
   volumeMode: Block
   storageClassName: local-storage

tests/integration/kubernetes/runtimeclass_workloads/nvidia-nim-llama-3-1-8b-instruct-tee.yaml.in

@@ -69,14 +69,20 @@ spec:
       limits:
         nvidia.com/pgpu: "1"
         cpu: "16"
-        memory: "64Gi"
+        memory: "48Gi"
     volumeMounts:
       - name: nim-trusted-cache
         mountPath: /opt/nim/.cache
+    volumeDevices:
+      - devicePath: /dev/trusted_store
+        name: trusted-storage
   volumes:
   - name: nim-trusted-cache
     emptyDir:
       sizeLimit: 64Gi
+  - name: trusted-storage
+    persistentVolumeClaim:
+      claimName: trusted-pvc-instruct
 ---
 apiVersion: v1
 kind: Secret

tests/integration/kubernetes/runtimeclass_workloads/nvidia-nim-llama-3-2-1b-instruct-service-tee.yaml.in

@@ -0,0 +1,54 @@
+# Copyright (c) 2026 NVIDIA Corporation
+#
+# SPDX-License-Identifier: Apache-2.0
+#
+---
+apiVersion: apps.nvidia.com/v1alpha1
+kind: NIMService
+metadata:
+  name: ${NIM_SERVICE_NAME}
+spec:
+  image:
+    repository: nvcr.io/nim/meta/llama-3.2-1b-instruct
+    tag: "1.12.0"
+    pullPolicy: IfNotPresent
+    pullSecrets:
+      - ngc-secret-llama-3-2-1b
+  authSecret: ngc-api-key-sealed-llama-3-2-1b
+  # /dev/trusted_store container image layer storage feature cannot be selected,
+  # but storage.emptyDir selects container data storage feature.
+  storage:
+    emptyDir:
+      sizeLimit: 10Gi
+  replicas: 1
+  resources:
+    limits:
+      nvidia.com/pgpu: "1"
+      cpu: "8"
+      memory: "56Gi"
+  expose:
+    service:
+      type: ClusterIP
+      port: 8000
+  runtimeClassName: kata
+  userID: 1000
+  groupID: 1000
+  annotations:
+    io.katacontainers.config.hypervisor.kernel_params: "agent.guest_components_procs=confidential-data-hub agent.aa_kbc_params=cc_kbc::${CC_KBS_ADDR}"
+    io.katacontainers.config.hypervisor.cc_init_data: "${NIM_INITDATA_BASE64}"
+---
+apiVersion: v1
+kind: Secret
+metadata:
+  name: ngc-secret-llama-3-2-1b
+type: kubernetes.io/dockerconfigjson
+data:
+  .dockerconfigjson: ${DOCKER_CONFIG_JSON}
+---
+apiVersion: v1
+kind: Secret
+metadata:
+  name: ngc-api-key-sealed-llama-3-2-1b
+type: Opaque
+data:
+  NGC_API_KEY: "${NGC_API_KEY_SEALED_SECRET_LLAMA_3_2_1B_BASE64}"

tests/integration/kubernetes/runtimeclass_workloads/nvidia-nim-llama-3-2-1b-instruct-service.yaml.in

@@ -0,0 +1,48 @@
+# Copyright (c) 2026 NVIDIA Corporation
+#
+# SPDX-License-Identifier: Apache-2.0
+#
+---
+apiVersion: apps.nvidia.com/v1alpha1
+kind: NIMService
+metadata:
+  name: ${NIM_SERVICE_NAME}
+spec:
+  image:
+    repository: nvcr.io/nim/meta/llama-3.2-1b-instruct
+    tag: "1.12.0"
+    pullPolicy: IfNotPresent
+    pullSecrets:
+      - ngc-secret-llama-3-2-1b
+  authSecret: ngc-api-key-llama-3-2-1b
+  storage:
+    hostPath: "${LOCAL_NIM_CACHE_LLAMA_3_2_1B}"
+  replicas: 1
+  resources:
+    limits:
+      nvidia.com/pgpu: "1"
+      cpu: "8"
+      memory: "16Gi"
+  expose:
+    service:
+      type: ClusterIP
+      port: 8000
+  runtimeClassName: kata
+  userID: 1000
+  groupID: 1000
+---
+apiVersion: v1
+kind: Secret
+metadata:
+  name: ngc-secret-llama-3-2-1b
+type: kubernetes.io/dockerconfigjson
+data:
+  .dockerconfigjson: ${DOCKER_CONFIG_JSON}
+---
+apiVersion: v1
+kind: Secret
+metadata:
+  name: ngc-api-key-llama-3-2-1b
+type: Opaque
+data:
+  NGC_API_KEY: "${NGC_API_KEY_BASE64}"

tests/integration/kubernetes/runtimeclass_workloads/nvidia-nim-llama-3-2-nv-embedqa-1b-v2-tee.yaml.in

@@ -83,10 +83,16 @@ spec:
     volumeMounts:
       - name: nim-trusted-cache
         mountPath: /opt/nim/.cache
+    volumeDevices:
+      - devicePath: /dev/trusted_store
+        name: trusted-storage
   volumes:
   - name: nim-trusted-cache
     emptyDir:
       sizeLimit: 40Gi
+  - name: trusted-storage
+    persistentVolumeClaim:
+      claimName: trusted-pvc-embedqa
 ---
 apiVersion: v1
 kind: Secret

tools/packaging/kata-deploy/binary/src/config.rs

@@ -155,6 +155,7 @@ pub struct Config {
     pub containerd_conf_file: String,
     pub containerd_conf_file_backup: String,
     pub containerd_drop_in_conf_file: String,
+    pub daemonset_name: String,
     pub custom_runtimes_enabled: bool,
     pub custom_runtimes: Vec<CustomRuntime>,
 }
@@ -169,6 +170,12 @@ impl Config {
             return Err(anyhow::anyhow!("NODE_NAME must not be empty"));
         }
 
+        let daemonset_name = env::var("DAEMONSET_NAME")
+            .ok()
+            .map(|v| v.trim().to_string())
+            .filter(|v| !v.is_empty())
+            .unwrap_or_else(|| "kata-deploy".to_string());
+
         let debug = env::var("DEBUG").unwrap_or_else(|_| "false".to_string()) == "true";
 
         // Parse shims - only use arch-specific variable
@@ -293,6 +300,7 @@ impl Config {
             containerd_conf_file,
             containerd_conf_file_backup,
             containerd_drop_in_conf_file,
+            daemonset_name,
             custom_runtimes_enabled,
             custom_runtimes,
         };

tools/packaging/kata-deploy/binary/src/k8s/client.rs

@@ -94,30 +94,41 @@ impl K8sClient {
         Ok(())
     }
 
-    pub async fn count_kata_deploy_daemonsets(&self) -> Result<usize> {
+    /// Returns whether a non-terminating DaemonSet with this exact name
+    /// exists in the current namespace. Used to decide whether this pod is
+    /// being restarted (true) or uninstalled (false).
+    pub async fn own_daemonset_exists(&self, daemonset_name: &str) -> Result<bool> {
+        use k8s_openapi::api::apps::v1::DaemonSet;
+        use kube::api::Api;
+
+        let ds_api: Api<DaemonSet> = Api::default_namespaced(self.client.clone());
+        match ds_api.get_opt(daemonset_name).await? {
+            Some(ds) => Ok(ds.metadata.deletion_timestamp.is_none()),
+            None => Ok(false),
+        }
+    }
+
+    /// Returns how many non-terminating DaemonSets across all namespaces
+    /// have a name containing "kata-deploy". Used to decide whether shared
+    /// node-level resources (node label, CRI restart) should be cleaned up:
+    /// they are only safe to remove when no kata-deploy instance remains
+    /// on the cluster.
+    pub async fn count_any_kata_deploy_daemonsets(&self) -> Result<usize> {
         use k8s_openapi::api::apps::v1::DaemonSet;
         use kube::api::{Api, ListParams};
 
-        let ds_api: Api<DaemonSet> = Api::default_namespaced(self.client.clone());
-        let lp = ListParams::default();
-        let daemonsets = ds_api.list(&lp).await?;
+        let ds_api: Api<DaemonSet> = Api::all(self.client.clone());
+        let daemonsets = ds_api.list(&ListParams::default()).await?;
 
-        // Note: We use client-side filtering here because Kubernetes field selectors
-        // don't support "contains" operations - they only support exact matches and comparisons.
-        // Filtering by name containing "kata-deploy" requires client-side processing.
-        // Exclude DaemonSets that are terminating (have deletion_timestamp) so that when our
-        // DaemonSet pod runs cleanup on SIGTERM during uninstall, we count 0 and remove the label.
         let count = daemonsets
             .iter()
             .filter(|ds| {
-                if ds.metadata.deletion_timestamp.is_some() {
-                    return false;
-                }
-                ds.metadata
-                    .name
-                    .as_ref()
-                    .map(|n| n.contains("kata-deploy"))
-                    .unwrap_or(false)
+                ds.metadata.deletion_timestamp.is_none()
+                    && ds
+                        .metadata
+                        .name
+                        .as_ref()
+                        .is_some_and(|n| n.contains("kata-deploy"))
             })
             .count();
 
@@ -584,9 +595,14 @@ pub async fn label_node(
     client.label_node(label_key, label_value, overwrite).await
 }
 
-pub async fn count_kata_deploy_daemonsets(config: &Config) -> Result<usize> {
+pub async fn own_daemonset_exists(config: &Config) -> Result<bool> {
     let client = K8sClient::new(&config.node_name).await?;
-    client.count_kata_deploy_daemonsets().await
+    client.own_daemonset_exists(&config.daemonset_name).await
+}
+
+pub async fn count_any_kata_deploy_daemonsets(config: &Config) -> Result<usize> {
+    let client = K8sClient::new(&config.node_name).await?;
+    client.count_any_kata_deploy_daemonsets().await
 }
 
 pub async fn crd_exists(config: &Config, crd_name: &str) -> Result<bool> {

tools/packaging/kata-deploy/binary/src/main.rs

@@ -236,19 +236,29 @@ async fn install(config: &config::Config, runtime: &str) -> Result<()> {
 async fn cleanup(config: &config::Config, runtime: &str) -> Result<()> {
     info!("Cleaning up Kata Containers");
 
-    info!("Counting kata-deploy daemonsets");
-    let kata_deploy_installations = k8s::count_kata_deploy_daemonsets(config).await?;
+    // Step 1: Check if THIS pod's owning DaemonSet still exists.
+    // If it does, this is a pod restart (rolling update, label change, etc.),
+    // not an uninstall — skip everything so running kata pods are not disrupted.
     info!(
-        "Found {} kata-deploy daemonset(s)",
-        kata_deploy_installations
+        "Checking if DaemonSet '{}' still exists",
+        config.daemonset_name
     );
-
-    if kata_deploy_installations == 0 {
-        info!("Removing kata-runtime label from node");
-        k8s::label_node(config, "katacontainers.io/kata-runtime", None, false).await?;
-        info!("Successfully removed kata-runtime label");
+    if k8s::own_daemonset_exists(config).await? {
+        info!(
+            "DaemonSet '{}' still exists, \
+             skipping all cleanup to avoid disrupting running kata pods",
+            config.daemonset_name
+        );
+        return Ok(());
     }
 
+    // Step 2: Our DaemonSet is gone (uninstall). Perform instance-specific
+    // cleanup: snapshotters, CRI config, and artifacts for this instance.
+    info!(
+        "DaemonSet '{}' not found, proceeding with instance cleanup",
+        config.daemonset_name
+    );
+
     match config.experimental_setup_snapshotter.as_ref() {
         Some(snapshotters) => {
             for snapshotter in snapshotters {
@@ -270,6 +280,25 @@ async fn cleanup(config: &config::Config, runtime: &str) -> Result<()> {
     artifacts::remove_artifacts(config).await?;
     info!("Successfully removed kata artifacts");
 
+    // Step 3: Check if ANY other kata-deploy DaemonSets still exist.
+    // Shared resources (node label, CRI restart) are only safe to touch
+    // when no other kata-deploy instance remains.
+    let other_ds_count = k8s::count_any_kata_deploy_daemonsets(config).await?;
+    if other_ds_count > 0 {
+        info!(
+            "{} other kata-deploy DaemonSet(s) still exist, \
+             skipping node label removal and CRI restart",
+            other_ds_count
+        );
+        return Ok(());
+    }
+
+    info!("No other kata-deploy DaemonSets found, performing full shared cleanup");
+
+    info!("Removing kata-runtime label from node");
+    k8s::label_node(config, "katacontainers.io/kata-runtime", None, false).await?;
+    info!("Successfully removed kata-runtime label");
+
     // Restart the CRI runtime last. On k3s/rke2 this restarts the entire
     // server process, which kills this (terminating) pod. By doing it after
     // all other cleanup, we ensure config and artifacts are already gone.

tools/packaging/kata-deploy/binary/src/runtime/manager.rs

@@ -51,18 +51,19 @@ pub async fn get_container_runtime(config: &Config) -> Result<String> {
         return Ok("crio".to_string());
     }
 
-    if runtime_version.contains("containerd") && runtime_version.contains("-k3s") {
-        // Check systemd services (ignore errors - service might not exist)
-        let _ = utils::host_systemctl(&["is-active", "--quiet", "rke2-agent"]);
-        if utils::host_systemctl(&["is-active", "--quiet", "rke2-agent"]).is_ok() {
-            return Ok("rke2-agent".to_string());
-        }
-        if utils::host_systemctl(&["is-active", "--quiet", "rke2-server"]).is_ok() {
-            return Ok("rke2-server".to_string());
-        }
-        if utils::host_systemctl(&["is-active", "--quiet", "k3s-agent"]).is_ok() {
-            return Ok("k3s-agent".to_string());
-        }
+    // Detect k3s/rke2 via systemd services rather than the containerd version
+    // string, which no longer reliably contains "k3s" in newer releases
+    // (e.g. "containerd://2.2.2-bd1.34").
+    if utils::host_systemctl(&["is-active", "--quiet", "rke2-agent"]).is_ok() {
+        return Ok("rke2-agent".to_string());
+    }
+    if utils::host_systemctl(&["is-active", "--quiet", "rke2-server"]).is_ok() {
+        return Ok("rke2-server".to_string());
+    }
+    if utils::host_systemctl(&["is-active", "--quiet", "k3s-agent"]).is_ok() {
+        return Ok("k3s-agent".to_string());
+    }
+    if utils::host_systemctl(&["is-active", "--quiet", "k3s"]).is_ok() {
         return Ok("k3s".to_string());
     }
 
@@ -83,7 +84,7 @@ pub async fn get_container_runtime(config: &Config) -> Result<String> {
     Ok(runtime)
 }
 
-/// Returns true if containerRuntimeVersion (e.g. "containerd://2.1.5-k3s1") indicates
+/// Returns true if containerRuntimeVersion (e.g. "containerd://2.1.5-k3s1", "containerd://2.2.2-bd1.34") indicates
 /// containerd 2.x or newer, false for 1.x or unparseable. Used for drop-in support
 /// and for K3s/RKE2 template selection (config-v3.toml.tmpl vs config.toml.tmpl).
 pub fn containerd_version_is_2_or_newer(runtime_version: &str) -> bool {
@@ -191,6 +192,7 @@ mod tests {
     #[case("containerd://2.0.0", true)]
     #[case("containerd://2.1.5", true)]
     #[case("containerd://2.1.5-k3s1", true)]
+    #[case("containerd://2.2.2-bd1.34", true)]
     #[case("containerd://2.2.0", true)]
     #[case("containerd://2.3.1", true)]
     #[case("containerd://2.0.0-rc.1", true)]

tools/packaging/kata-deploy/helm-chart/kata-deploy/templates/kata-deploy.yaml

@@ -143,6 +143,13 @@ spec:
           valueFrom:
             fieldRef:
               fieldPath: spec.nodeName
+{{- if .Values.env.multiInstallSuffix }}
+        - name: DAEMONSET_NAME
+          value: {{ printf "%s-%s" .Chart.Name .Values.env.multiInstallSuffix | quote }}
+{{- else }}
+        - name: DAEMONSET_NAME
+          value: {{ .Chart.Name | quote }}
+{{- end }}
         - name: DEBUG
           value: {{ include "kata-deploy.getDebug" . | quote }}
         {{- $shimsAmd64 := include "kata-deploy.getEnabledShimsForArch" (dict "root" . "arch" "amd64") | trim -}}

tools/packaging/kata-deploy/nydus-snapshotter/nydus-snapshotter.service

@@ -5,6 +5,8 @@ Before=containerd.service
 
 [Service]
 ExecStart=@CONTAINERD_NYDUS_GRPC_BINARY@ --config @CONFIG_GUEST_PULLING@ --log-to-stdout
+Restart=always
+RestartSec=5
 
 [Install]
-RequiredBy=containerd.service
+WantedBy=containerd.service