wbuchwalter

2016-12-12 10:06:01 upgrade tar:amd64 1.28-2.1 1.28-2.1ubuntu0.1
2016-12-12 10:06:01 status half-configured tar:amd64 1.28-2.1
2016-12-12 10:06:01 status unpacked tar:amd64 1.28-2.1
2016-12-12 10:06:01 status half-installed tar:amd64 1.28-2.1
2016-12-12 10:06:01 status triggers-pending man-db:amd64 2.7.5-1
2016-12-12 10:06:01 status triggers-pending mime-support:all 3.59ubuntu1
2016-12-12 10:06:02 status half-installed tar:amd64 1.28-2.1
2016-12-12 10:06:02 status unpacked tar:amd64 1.28-2.1ubuntu0.1
2016-12-12 10:06:02 status unpacked tar:amd64 1.28-2.1ubuntu0.1
2016-12-12 10:06:02 trigproc man-db:amd64 2.7.5-1 <none>

wbuchwalter

Start-Date: 2016-12-12  10:05:56
Commandline: /usr/bin/unattended-upgrade
Install: linux-headers-4.4.0-53-generic:amd64 (4.4.0-53.74, automatic), linux-cloud-tools-4.4.0-53:amd64 (4.4.0-53.74, automatic), linux-cloud-tools-4.4.0-53-generic:amd64 (4.4.0-53.74, automatic), linux-headers-4.4.0-53:amd64 (4.4.0-53.74, automatic)
Upgrade: lxc-common:amd64 (2.0.5-0ubuntu1~ubuntu16.04.2, 2.0.5-0ubuntu1~ubuntu16.04.3), linux-headers-generic:amd64 (4.4.0.47.50, 4.4.0.53.56), vim-common:amd64 (2:7.4.1689-3ubuntu1.1, 2:7.4.1689-3ubuntu1.2), python2.7-minimal:amd64 (2.7.12-1~16.04, 2.7.12-1ubuntu0~16.04.1), libpython2.7:amd64 (2.7.12-1~16.04, 2.7.12-1ubuntu0~16.04.1), python2.7:amd64 (2.7.12-1~16.04, 2.7.12-1ubuntu0~16.04.1), libpython3.5:amd64 (3.5.2-2~16.04, 3.5.2-2ubuntu0~16.04.1), python3.5:amd64 (3.5.2-2~16.04, 3.5.2-2ubuntu0~16.04.1), python3.5-minimal:amd64 (3.5.2-2~16.04, 3.5.2-2ubuntu0~16.04.1), linux-cloud-tools-common:amd64 (4.4.0-47.68, 4.4.0-53.74), liblxc1:amd64 (2.0.5-0ubuntu1~ubuntu16.04.2, 2.0.5-0ubuntu1~

wbuchwalter

{
    "coinbase": "0x0",
    "difficulty": "0x20000",
    "extraData": "0x0",
    "gasLimit": "0x2625A0",
    "mixhash": "0x0",
    "nonce": "0x0216",
    "parentHash": "0x0",
    "timestamp": "0x00",
    "alloc": {

wbuchwalter

$ kubelet \  
--api-servers="https://{{{masterPrivateIp}}}:443" \	  
--experimental-nvidia-gpus 1 \  
--kubeconfig=/var/lib/kubelet/kubeconfig \	  
--address=0.0.0.0 \		         
--allow-privileged=true  

wbuchwalter

{
    "parameters": {
        "setLinuxConfigurationForVMCreate": {
            "type": "int",
            "allowedValues": [
                0,
                1
            ],
            "metadata": {
                "description": "This setting controls whether Linux configuration with SSH Key is passed in VM PUT Payload.  Defaults to 1.  If SSH Key is blank, this must be set to 0."

wbuchwalter

{
  "apiVersion": "vlabs",
  "properties": {
    "orchestratorProfile": {
      "orchestratorType": "Kubernetes"
    },
    "masterProfile": {
      "count": 1,
      "dnsPrefix": "",
      "vmSize": "Standard_D2_v2"

wbuchwalter

"agentPoolProfiles": [
  {
    "name": "agentpool1",
    "count": 3,
    "vmSize": "Standard_D2_v2",
    "availabilityProfile": "AvailabilitySet"
  }
]

wbuchwalter

{
  "apiVersion": "vlabs",
  "properties": {
    "orchestratorProfile": {
      "orchestratorType": "Kubernetes"
    },
    "masterProfile": {
      "count": 1,
      "dnsPrefix": "",
      "vmSize": "Standard_D2_v2"

wbuchwalter

1. Create k8s cluster

• Use my fork of ACS-Engine on the k8s-gpu-flag branch
• Launch and build ACS-Engine, and customize the kubernetes example template to have NC6 VMs (example
• Create RG: az group create --name k8s --location southcentralus (or other region with GPU)
• Deploy the generated template az group deployment create --ressource-group k8s --template-file azuredeploy.json --parameters @azuredeploy.pamareters.json

2. Install drivers

• Either SSH into each node and run this script : install-nvidia-driver.sh

wbuchwalter

#!/bin/bash
az account clear
python main.py --resource-group acs --container-service-name containerservice-acs \
 --service-principal-app-id 'xxx' --service-principal-secret 'xxx' \
 --service-principal-tenant-id 'xxx' \
 -vvv --kubeconfig /root/.kube/config --over-provision 0 --debug