images including kinds of RDMA tools for debugging
{cuda version of baseImage}-{commit hash of directory}
the version of chart is '{x of cudaVersion}-{y of cudaVersion}-{custom}'.
For example of chart version 'v12.5.0', '12.5' represents cuda '12.5.x' in the base image, the last '.0' represents any changes of the chart
tag the code and the CI will automatically release a chart
git tag rdma-tools-vXX.YY.ZZ
git push --tagshelm repo add spiderchart https://spidernet-io.github.io/charts
helm repo update spiderchart
helm search repo rdma-tools
# run daemonset on worker1 and worker2
cat <<EOF > values.yaml
# for china user , it could add these to use a domestic registry
#image:
# registry: ghcr.m.daocloud.io
# just run daemonset in nodes 'worker1' and 'worker2'
affinity:
nodeAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
nodeSelectorTerms:
- matchExpressions:
- key: kubernetes.io/hostname
operator: In
values:
- worker1
- worker2
# sriov interfaces
extraAnnotations:
v1.multus-cni.io/default-network: spiderpool/calico
k8s.v1.cni.cncf.io/networks: |-
[{"name":"gpu1-sriov","namespace":"spiderpool"},
{"name":"gpu2-sriov","namespace":"spiderpool"}]
# sriov resource
resources:
limits:
spidernet.io/gpu1sriov: 1
spidernet.io/gpu2sriov: 1
# nvidia.com/gpu: 1
#securityContext:
# # required by gdrcopy test
# privileged: true
# capabilities:
# add: [ "IPC_LOCK" ]
EOF
# for China user, add `--set image.registry=ghcr.m.daocloud.io`
helm install rdma-tools spiderchart/rdma-tools -f ./values.yaml
os: ubuntu22.04
| tools | version | updated time |
|---|---|---|
| cuda | 12.5.1 | 2024.7.30 |
| nccl | 2.22.3 | 2024.7.30 |
| nccl-test | v2.13.10 | 2024.7.30 |
| hpc-x | v2.19 | 2024.7.30 |
| gdrcopy | 1366e20d140c5638fcaa6c72b373ac69f7ab2532 | 2024.7.30 |
| perftest | 24.04.0-0.41 | 2024.7.30 |
| Bandwidthtest | v12.5 | 2024.7.30 |
| nvbandwidth | v0.5 | 2024.8.14 |