Kubernetes
Open-Source GPU Virtualization Projects on Kubernetes
· ☕ 5 min read
1. k8s-device-plugin https://github.com/NVIDIA/k8s-device-plugin is the official Kubernetes device plugin from NVIDIA, used to manage and allocate NVIDIA GPU resources in a Kubernetes cluster. By interacting with kubelet, k8s-device-plugin automatically discovers and registers GPU devices and exposes them to the Kubernetes scheduler as resources. It supports many GPU models and is able to handle GPU partitioning and sharing.

FoundationDB Data Backup and Restore
· ☕ 2 min read
1. Set Environment Variables Set the foundationdb cluster name and version 1 2 3 4 5 export CLUSTER_NAME= export NAMESPACE= export DEPLOY_NAME=${CLUSTER_NAME}-deployment export SECRET_NAME=fdb-backup-secret export VERSION=7.1.26 Set the S3 backup bucket and credentials 1 2 3 4 5 export BUCKET= export HOST=s3.us-west-2.amazonaws.com export AK=x export SK=x export S3_BACKUP_URI="blobstore://${AK}:${SK}@${HOST}/fdb-backup?bucket=${BUCKET}&sc=0" 2. Create the Credentials 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 cat <<EOF | kubectl apply -f - apiVersion: v1 kind: Secret metadata: name: ${SECRET_NAME} namespace: ${NAMESPACE} type: Opaque stringData: credentials: | { "accounts": { "${BUCKET}@${HOST}": { "access_key": "${AK}", "secret": "${SK}", "endpoint": "${HOST}" } } } EOF 3.

kube-proxy Failure Prevents Pods on a Node from Accessing a Service
· ☕ 5 min read
1. Problem Description Related Pods 1 2 3 4 5 6 kubectl -n istio-system get pod -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES istiod-647c7c9d95-7n7n6 1/1 Running 0 77m 10.244.173.51 docs-ai-a800-4 <none> <none> istiod-647c7c9d95-k6l88 1/1 Running 0 30m 10.244.210.160 ai-a40-2 <none> <none> istiod-647c7c9d95-pj82r 1/1 Running 0 51m 10.

Deploying the DeepSeek 3FS Storage System in Containers
· ☕ 12 min read
1. Deployment Plan Before starting the containerized deployment, a few requirements: To simplify delivery, only one image is needed For reliability, deploy as many replicas as possible Start different services through different parameters Inject configuration through environment variables and render it into configuration files Below is the deployment plan for DeepSeek 3FS:

Using Calico to Limit Pod Bandwidth
· ☕ 5 min read
1. Testing the Full Bandwidth Start the iperf3 server on the target host 1 2 3 4 5 iperf3 -s ----------------------------------------------------------- Server listening on 5201 ----------------------------------------------------------- Test from the client host 1 iperf3 -c x.x.x.x -p 5201 -t 10 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Connecting to host x.