fix(examples): update Triton example for CUDA 13 + fix libdcgm copy (DYN-3697) by nealvaidya · Pull Request #12577 · ai-dynamo/dynamo

nealvaidya · 2026-08-03T16:54:15Z

Summary

Bumps default TRITON_SERVER_IMAGE from tritonserver:25.01-py3 (CUDA 12.8) to tritonserver:25.10-py3 (CUDA 13.0.2) to match the CUDA 13-only Dynamo base image. The old image's triton_bindings linked against libdcgm.so.3 / libcudart.so.12, both absent in the CUDA 13 environment.
Fixes the DCGM library copy: Triton 25.10 ships libdcgm.so.4 under /usr/lib/x86_64-linux-gnu (merged-usr convention) rather than /lib/x86_64-linux-gnu. Updates the bind-mount source path accordingly.
Adds USER root before the find … cp step so the dynamo-base non-root user does not cause a permission-denied failure when writing to /lib/x86_64-linux-gnu/.
Adds a build-time python3 -c "import tritonserver" smoke check so future ABI or library mismatches fail the image build rather than producing a silently broken image.

Validation

Built dynamo-triton:latest from scratch on a host with 5× H100 80GB:

docker build --build-arg DYNAMO_BASE_IMAGE=dynamo-base:latest \
  -t dynamo-triton:latest examples/backends/tritonserver/

All 11 build steps passed, including the import tritonserver smoke check.

Ran the identity server and client end-to-end:

docker run --rm -d --gpus all --network host \
  -e DYN_HTTP_PORT=8777 --name dynamo-triton-test \
  dynamo-triton:latest /workspace/launch/identity.sh

docker exec dynamo-triton-test \
  python3 /workspace/src/client.py --port 8777 --iterations 3

✓ Identity verification passed   # iteration 1
✓ Identity verification passed   # iteration 2
✓ Identity verification passed   # iteration 3

Note: DYN_HTTP_PORT override needed when port 8000 is already in use on the host.

🤖 Generated with Claude Code

Summary by CodeRabbit

Documentation
- Updated Triton server container build and deployment instructions for CUDA 13.0.
- Clarified the default server image and corrected the container launch command.
Bug Fixes
- Improved Triton container setup by reliably including required GPU libraries.
- Added build-time validation to confirm the Triton server package is available.

…-3697) Bump default TRITON_SERVER_IMAGE from tritonserver:25.01-py3 (CUDA 12.8) to tritonserver:25.10-py3 (CUDA 13.0.2) to match the CUDA 13-only Dynamo base image produced by container/render.py. The old image's triton_bindings linked against libcudart.so.12, which is absent in the CUDA 13 environment, causing the worker to fail on import before model registration. Also make the DCGM library copy tolerant of releases where /usr/local/dcgm is absent, and add a build-time `import tritonserver` check so future ABI mismatches fail the image build rather than producing a silently broken image. Signed-off-by: Neal Vaidya <nealv@nvidia.com>

) Triton 25.10 ships libdcgm.so.4 under /usr/lib/x86_64-linux-gnu (the merged-usr convention) rather than /lib/x86_64-linux-gnu. Update the bind-mount source accordingly and add USER root before the copy so the dynamo-base non-root user does not cause a permission-denied failure. Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com> Signed-off-by: Neal Vaidya <nealv@nvidia.com>

devin-ai-integration

Devin Review found 2 potential issues.

coderabbitai · 2026-08-03T16:57:44Z

Walkthrough

Changes

Triton container workflow

Layer / File(s)	Summary
Triton image build and validation `examples/backends/tritonserver/Dockerfile`	The Dockerfile uses Triton 25.10, copies matching DCGM libraries, and validates the `tritonserver` import during the build.
Container deployment instructions `examples/backends/tritonserver/README.md`	The README updates CUDA, image-build, default-image, and container-launch instructions.

Estimated code review effort: 2 (Simple) | ~10 minutes

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	The description explains the changes and validation, but it omits the required Related Issues section and reviewer starting point.	Add the required Related Issues section and identify the files or sections where the reviewer should start.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the Triton example update for CUDA 13 and the DCGM library copy fix.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{Comment @coderabbitai help to get the list of available commands.}

coderabbitai

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@examples/backends/tritonserver/Dockerfile`:
- Around line 31-32: Update the comment immediately above the `RUN python3 -c
"import tritonserver"` command to state that it validates the tritonserver
Python package import at build time; do not describe it as CUDA ABI
compatibility validation.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

Push a commit to this branch (recommended)
Create a new PR with the fixes

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 2e9c9786-c687-49f0-96aa-6d1487bbc1a4

📥 Commits

Reviewing files that changed from the base of the PR and between 58d289a and 9208022.

📒 Files selected for processing (2)

examples/backends/tritonserver/Dockerfile
examples/backends/tritonserver/README.md

github-actions · 2026-08-03T16:58:57Z

🌿 Fern Docs Preview: https://nvidia-preview-019fc890-2a82-735b-a0e5-3e1abacd0ca8.docs.buildwithfern.com/dynamo/dev

…YN-3697) (#12577) Signed-off-by: Neal Vaidya <nealv@nvidia.com> Co-authored-by: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com> (cherry picked from commit ba294ab) Signed-off-by: Dan Gil <dagil@nvidia.com>

… (#12604) Signed-off-by: Neal Vaidya <nealv@nvidia.com> Signed-off-by: Dan Gil <dagil@nvidia.com> Co-authored-by: Neal Vaidya <nealv@nvidia.com> Co-authored-by: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

dyn-3691-extract-shared-target-pid-cuda-customstorage-operation-layer * 'main' of https://github.com/ai-dynamo/dynamo: (50 commits) docs(cli): correct removed vLLM prefill-worker flag reference (#12581) docs(operator): reserve webhook Ignore for emergencies (#12563) ci(docs): make previews and checks match what actually publishes (#12339) refactor(vllm): organize custom encoder modules (#12416) feat(llm): Select reasoning output field via env var (#11464) feat(runtime): add TLS support to TCP request plane (#10921) fix: convert conditional disagg sglang warning to httperror 400 (#12578) feat(operator): add runtime feature gates (#12421) refactor(runtime): extract PushRouter transport seam behind StreamingDispatch trait (#12447) feat(replay): add deterministic canonical offline reports (#12363) build: bump ModelExpress to 0.5.0(OPS-7978) (#12455) fix(mocker): use logical KV tokens for decode timing (#12583) fix(examples): update Triton example for CUDA 13 + fix libdcgm copy (DYN-3697) (#12577) refactor(operator): implement composition-first DGD reconciliation (#12283) feat(frontend): add basetenkenizer backend (#12376) fix(profiler): configure rapid mocker without planner (#12573) docs(vllm): correct worker-role flags and document --kv-transfer-config (#12568) ci: add Kubernetes deploy test to nightly (#12090) fix(container): reuse pinned protoc in runtime image (#12535) feat(self-host): flip DYN_SELF_HOST_METADATA default to ON (gh-8749) (#11417) ... Signed-off-by: Hannah Zhang <hannahz@nvidia.com>

nealvaidya and others added 2 commits July 31, 2026 14:44

nealvaidya requested review from a team as code owners August 3, 2026 16:54

pull-request-size Bot added the size/S label Aug 3, 2026

github-actions Bot added fix documentation Improvements or additions to documentation labels Aug 3, 2026

devin-ai-integration Bot reviewed Aug 3, 2026

View reviewed changes

Comment thread examples/backends/tritonserver/Dockerfile

Comment thread examples/backends/tritonserver/README.md

coderabbitai Bot reviewed Aug 3, 2026

View reviewed changes

Comment thread examples/backends/tritonserver/Dockerfile

dynamo-review-agent Bot approved these changes Aug 3, 2026

View reviewed changes

dagil-nvidia mentioned this pull request Aug 3, 2026

fix(examples): bump Triton example image to a CUDA-13 tag #12569

Closed

1 task

dagil-nvidia approved these changes Aug 3, 2026

View reviewed changes

dagil-nvidia merged commit ba294ab into main Aug 3, 2026
98 checks passed

dagil-nvidia deleted the nealv/dyn-3697 branch August 3, 2026 22:19

dagil-nvidia mentioned this pull request Aug 3, 2026

fix(examples): cherry-pick Triton CUDA 13 image + libdcgm copy (#12577) #12604

Merged

Navigation Menu

Uh oh!

fix(examples): update Triton example for CUDA 13 + fix libdcgm copy (DYN-3697) - #12577

fix(examples): update Triton example for CUDA 13 + fix libdcgm copy (DYN-3697)#12577
dagil-nvidia merged 2 commits into
mainfrom
nealv/dyn-3697

nealvaidya commented Aug 3, 2026 •

edited by coderabbitai Bot

Loading

Uh oh!

devin-ai-integration Bot left a comment

Uh oh!

Uh oh!

Uh oh!

coderabbitai Bot commented Aug 3, 2026

❌ Failed checks (1 warning)

Uh oh!

coderabbitai Bot left a comment

Uh oh!

Uh oh!

github-actions Bot commented Aug 3, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

nealvaidya commented Aug 3, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Validation

Summary by CodeRabbit

Uh oh!

devin-ai-integration Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

coderabbitai Bot commented Aug 3, 2026

Walkthrough

Changes

❌ Failed checks (1 warning)

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

github-actions Bot commented Aug 3, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

nealvaidya commented Aug 3, 2026 •

edited by coderabbitai Bot

Loading