@tduncan / Nvidia-Optimization-Stack-VLLM
Code Commits Issues Pulls Cloud Agent CI/CD
main
main
yaml 130 lines 4.41 KB
Raw
1 group: Abuild
2 steps:
3 - label: ":docker: Build image"
4 key: image-build
5 depends_on: []
6 timeout_in_minutes: 600
7 commands:
8 - if [[ "$BUILDKITE_BRANCH" == "main" ]]; then .buildkite/image_build/image_build.sh $REGISTRY $REPO $BUILDKITE_COMMIT $BRANCH $IMAGE_TAG $IMAGE_TAG_LATEST; else .buildkite/image_build/image_build.sh $REGISTRY $REPO $BUILDKITE_COMMIT $BRANCH $IMAGE_TAG; fi
9 retry:
10 automatic:
11 - exit_status: -1 # Agent was lost
12 limit: 2
13 - exit_status: -10 # Agent was lost
14 limit: 2
15
16 - label: ":docker: :smoking: Non-root smoke tests"
17 key: image-build-smoke-test
18 depends_on:
19 - image-build
20 commands:
21 # Smoke 1: the default (root) image must still be importable
22 # under a non-root UID via `--user 2000:0`. Validates the `vllm` passwd
23 # entry + group-0-writable /home/vllm + uv path cleanup from #31959.
24 # Uses `import vllm` rather than `vllm serve --help` because the latter
25 # instantiates `VllmConfig` which requires a GPU attached to the
26 # container.
27 - docker run --rm --user 2000:0 --entrypoint python3 "$IMAGE_TAG" -c "import vllm; print(vllm.__version__)"
28 # Smoke 2: assert the non-root enabling invariants are baked
29 # into the image. Runs as UID 2000:0 via a shell so we can verify
30 # filesystem perms + passwd/group file state + wrapper presence without
31 # triggering vLLM's GPU-requiring config-init path. The opt-in
32 # `vllm-openai-nonroot` target adds only `USER vllm`, `WORKDIR
33 # /home/vllm`, and an `ENTRYPOINT` override on top of these invariants;
34 # its build correctness is reviewed at the Dockerfile level. Wrapper
35 # logic is covered separately by the pre-commit hook
36 # `test-nonroot-entrypoint` (see .pre-commit-config.yaml).
37 - |
38 docker run --rm --user 2000:0 --entrypoint /bin/sh "$IMAGE_TAG" -ec '
39 if ! getent passwd 2000 | grep -q ^vllm:; then
40 echo FAIL: UID 2000 != vllm
41 exit 1
42 fi
43 if ! id -gn 2>/dev/null | grep -qx root; then
44 echo FAIL: GID 0 not root group
45 exit 1
46 fi
47 touch /home/vllm/.smoke && rm /home/vllm/.smoke
48 touch /opt/uv/cache/.smoke && rm /opt/uv/cache/.smoke
49 if ! test -x /usr/local/bin/vllm-nonroot-entrypoint.sh; then
50 echo FAIL: wrapper missing
51 exit 1
52 fi
53 if ! test -w /etc/passwd; then
54 echo FAIL: /etc/passwd not group-writable
55 exit 1
56 fi
57 if ! test -w /etc/group; then
58 echo FAIL: /etc/group not group-writable
59 exit 1
60 fi
61 echo non-root invariants OK
62 '
63 retry:
64 automatic:
65 - exit_status: -1 # Agent was lost
66 limit: 2
67 - exit_status: -10 # Agent was lost
68 limit: 2
69
70 - label: ":docker: Build CPU image"
71 key: image-build-cpu
72 depends_on: []
73 commands:
74 - .buildkite/image_build/image_build_cpu.sh $REGISTRY $REPO $BUILDKITE_COMMIT
75 env:
76 DOCKER_BUILDKIT: "1"
77 retry:
78 automatic:
79 - exit_status: -1 # Agent was lost
80 limit: 2
81 - exit_status: -10 # Agent was lost
82 limit: 2
83
84 - label: ":docker: Build HPU image"
85 soft_fail: true
86 depends_on: []
87 key: image-build-hpu
88 commands:
89 - .buildkite/image_build/image_build_hpu.sh $REGISTRY $REPO $BUILDKITE_COMMIT
90 env:
91 DOCKER_BUILDKIT: "1"
92 retry:
93 automatic:
94 - exit_status: -1 # Agent was lost
95 limit: 2
96 - exit_status: -10 # Agent was lost
97 limit: 2
98
99 - label: ":docker: Build CPU arm64 image"
100 key: cpu-arm64-image-build
101 depends_on: []
102 optional: true
103 commands:
104 - .buildkite/image_build/image_build_cpu_arm64.sh $REGISTRY $REPO $BUILDKITE_COMMIT
105 env:
106 DOCKER_BUILDKIT: "1"
107 retry:
108 automatic:
109 - exit_status: -1 # Agent was lost
110 limit: 2
111 - exit_status: -10 # Agent was lost
112 limit: 2
113
114 - label: ":docker: Build arm64 image"
115 key: arm64-image-build
116 depends_on: []
117 source_file_dependencies:
118 - ".buildkite/image_build/image_build.yaml"
119 - ".buildkite/image_build/image_build_arm64.sh"
120 - "docker/Dockerfile"
121 commands:
122 - .buildkite/image_build/image_build_arm64.sh $REGISTRY $REPO $BUILDKITE_COMMIT
123 env:
124 DOCKER_BUILDKIT: "1"
125 retry:
126 automatic:
127 - exit_status: -1 # Agent was lost
128 limit: 2
129 - exit_status: -10 # Agent was lost
130 limit: 2