ci: 由 VM 任务启动本地 Docker daemon
test / shell (pull_request) Successful in 41s
test / python (pull_request) Successful in 1m29s
test / go (pull_request) Successful in 3m7s

This commit is contained in:
2026-09-21 12:00:55 +00:00
parent 51b940468e
commit fef7e5a214
2 changed files with 44 additions and 3 deletions
+39
View File
@@ -9,6 +9,45 @@ jobs:
kind: kind:
runs-on: [self-hosted, vm-dev] runs-on: [self-hosted, vm-dev]
steps: steps:
- name: Start job-local Docker
shell: bash
run: |
set -euo pipefail
sudo install -d /var/lib/docker
sudo truncate -s 20G /tmp/docker-data.img
sudo mkfs.ext4 -F /tmp/docker-data.img
sudo mount -o loop /tmp/docker-data.img /var/lib/docker
# A nested systemd needs a domain cgroup namespace. This is the
# cgroup v2 nesting initialization performed by the official DinD
# entrypoint, kept here because the shared runner image deliberately
# does not carry a second DinD-specific entrypoint.
if [[ -f /sys/fs/cgroup/cgroup.controllers ]]; then
sudo sh -c '
mkdir -p /sys/fs/cgroup/init
while read -r pid; do
printf "%s\n" "$pid" \
>/sys/fs/cgroup/init/cgroup.procs 2>/dev/null || true
done </sys/fs/cgroup/cgroup.procs
sed -e "s/ / +/g" -e "s/^/+/" \
/sys/fs/cgroup/cgroup.controllers \
>/sys/fs/cgroup/cgroup.subtree_control
'
fi
sudo nohup dockerd \
--host=unix:///var/run/docker.sock \
--storage-driver=overlay2 \
>/tmp/dockerd.log 2>&1 &
for _ in {1..60}; do
if docker info >/dev/null 2>&1; then
exit 0
fi
sleep 1
done
cat /tmp/dockerd.log
exit 1
- name: Verify Docker - name: Verify Docker
shell: bash shell: bash
run: | run: |
+5 -3
View File
@@ -40,9 +40,11 @@ UID attestation 的临时 Agent 失去父级。
`ci-vm` 使用 `kata-clh-runtime-rs`;`ci-pod` 使用默认 runc。两者都要求: `ci-vm` 使用 `kata-clh-runtime-rs`;`ci-pod` 使用默认 runc。两者都要求:
- runner 镜像包含 Gitea Runner、Node.js action userspace、SPIRE CLI 和 identity gate; - runner 镜像包含 Gitea Runner、Node.js action userspace、SPIRE CLI、Docker 工具和
- runner UID 2000,SPIRE Agent 与 privileged dockerd 使用不同 UID; identity gate;Pod 与 VM backend 使用同一个镜像;
- Docker socket 通过 group 2000 共享,Docker 数据仅存在于 sandbox emptyDir; - runner UID 2000;SPIRE Agent 独立运行,需要 Docker 的 workflow 通过 sudo 在
privileged executor 内启动 job-local daemon;
- Kata VM 中 Docker 数据使用 guest 内的 loop-backed ext4,并随 sandbox 一起删除;
- `self-hosted` 必须是所有 runner labels 的前缀; - `self-hosted` 必须是所有 runner labels 的前缀;
- ephemeral/once runner 完成一项任务后退出。 - ephemeral/once runner 完成一项任务后退出。