Repository navigation
Fix the existing a3-mega inferencing recipes from old fastrak versions - #169
Merged
Merged
Conversation
…harts - Upgrade A3 Mega FasTrak TCPXO plugin and RxDM daemon to LKG (nccl-v1.0.30-external: v1.0.18 / v1.0.24 / buildType 228) and mount at /usr/local/tcpxo across single-host and multi-host vLLM/SGLang Helm templates. - Set HF_HOME=/ssd on LocalSSD and bound gke-gcsfuse sidecar memory/cache (32Gi limit, max-size-mb:0) to prevent Exit Code 137 node OOMs during 671B weight downloads. - Align SGLang LeaderWorkerSet template networking (hostNetwork: true, ClusterFirstWithHostNet, GLOO_SOCKET_IFNAME=eth0) and symmetric leader/worker NCCL env vars with vLLM multi-host chart. - Remove duplicate Service selector keys in multi-host LWS Service templates and fix Llama-4 Maverick FP8 model ID and README instructions.
Chris113113
force-pushed
the
pirillo/fix_dsr1
branch
from
October 16, 2026 21:20
8caf3f9 to
b258909
Compare
- Fix vLLM multi-host LWS ray_init.sh path and executable permissions:
invoke `bash /workspace/ray_init.sh` in `lws-deployment.yaml`, mark
`src/docker/vllm/ray_init.sh` executable (`100755`), and ensure both
`/workspace/ray_init.sh` and `/vllm-workspace/ray_init.sh` exist in
`src/docker/vllm/vllm.Dockerfile`.
- Fix SGLang multi-host LWS bind address: pass `--host 0.0.0.0` and
`--port {{ $root.Values.sglang.service.ports.http }}` to
`python3 -m sglang.launch_server` in `sglang-inference/templates/lws-deployment.yaml`
so the server binds to all interfaces instead of `127.0.0.1:30000`,
allowing Kubelet readinessProbe and Kubernetes Service routing to succeed.
- Add `HF_HUB_ETAG_TIMEOUT=60` and `HF_HUB_DOWNLOAD_TIMEOUT=60` across
single-host vLLM, multi-host vLLM, and SGLang Helm templates to prevent
transient HuggingFace Xet CDN read timeouts during concurrent 8-rank
safetensors shard downloads.
Chris113113
marked this pull request as ready for review
October 19, 2026 16:51
joeywan-google
approved these changes
Oct 19, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This pull request fixes the GKE A3-Mega inference recipes and Helm charts for
deepseek-ai/DeepSeek-R1(vLLM and SGLang) andmeta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8(vLLM).Root Causes
Multi-node and single-node A3-Mega inference deployments failed during pod startup, weight download, or readiness probing for five reasons:
v1.0.9andv1.0.12) withNCCL_NET_GDR_LEVEL="SYS", which triggeredncclUnhandledCudaErrorandncclInternalErrorduring multi-node communicator initialization.sglang-inferenceLeaderWorkerSet template launchedsglang.launch_serverwithout--hostor--portflags, binding only to127.0.0.1:30000and causing Kubelet readiness probes and ClusterIP service traffic to fail withConnection refused.vllm-inference/multi-hostLeaderWorkerSet template invoked/vllm-workspace/ray_init.sh, whereassrc/docker/vllm/vllm.DockerfilesetsWORKDIR /workspaceand copiedray_init.shwithout executable permissions (exit code126).huggingface_hubread timeout.vllm/vllm-openai:v0.7.2(prior toLlama4ForConditionalGenerationsupport inv0.8.3), pointed to the removedsrc/helm-charts/a3mega/vllm-inferenceroot path instead ofsrc/helm-charts/a3mega/vllm-inference/single-host, and set--max-model-len=128000(exceeding single-node H100 80GB KV cache headroom after loading FP8 weights).Changes
The updates span the A3-Mega Helm templates, recipe values, and Dockerfiles:
ncclPluginImagetous-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.18andrxdmImagetous-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.24across all A3-Mega inference recipes (useHostPlugin: false,ncclBuildType: "12.4",NCCL_NET_GDR_LEVEL: "PIX").--host 0.0.0.0and--port {{ $root.Values.sglang.service.ports.http }}tosglang.launch_serverinsrc/helm-charts/a3mega/sglang-inference/templates/lws-deployment.yaml, configuretcpxo-daemonas a native sidecar container (restartPolicy: Always), and remove the unsupportedsubGroupPolicyTypefield.src/docker/vllm/ray_init.shexecutable (0755), populate both/workspace/ray_init.shand/vllm-workspace/ray_init.shinsrc/docker/vllm/vllm.Dockerfile, and invokebash /workspace/ray_init.shinsrc/helm-charts/a3mega/vllm-inference/multi-host/templates/lws-deployment.yaml.HF_HUB_ETAG_TIMEOUT="60"andHF_HUB_DOWNLOAD_TIMEOUT="60"acrosssingle-host,multi-host, andsglang-inferenceHelm templates, and quote containerimagefields sohelm lintsucceeds with default placeholder values.inference/a3mega/llama-4/vllm-serving-gketovllm/vllm-openai:v0.8.3, setmax-model-len: 32768, remove the unsupportedtrust-remote-code: trueentry fromvllm.serverArgs, and point the README tosrc/helm-charts/a3mega/vllm-inference/single-host.