From 96a7579322687f956d724ab2b9620f7bf3a088fa Mon Sep 17 00:00:00 2001
From: dkeven <82354774+dkeven@users.noreply.github.com>
Date: Mon, 6 Jan 2025 23:06:11 +0800
Subject: [PATCH] feat(installer): add gpu commands (#826)

* feat: add node selector

* feat(installer): install gpu driver & plugin by default

* fix: label bug

* fix: update installer

---------

Co-authored-by: liuyu <>
---
 build/installer/deploy/device-plugin.yaml       | 2 ++
 build/installer/deploy/nvidia-device-plugin.yml | 2 ++
 build/installer/deploy/scheduler.yaml           | 2 ++
 build/installer/install.ps1                     | 2 +-
 build/installer/install.sh                      | 2 +-
 5 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/build/installer/deploy/device-plugin.yaml b/build/installer/deploy/device-plugin.yaml
index ff15c514..8850aa69 100644
--- a/build/installer/deploy/device-plugin.yaml
+++ b/build/installer/deploy/device-plugin.yaml
@@ -28,6 +28,8 @@ spec:
     spec:
       runtimeClassName: nvidia # Explicitly request the runtime
       priorityClassName: system-node-critical
+      nodeSelector:
+        gpu.bytetrade.io/cuda-supported: 'true'
       initContainers:
       - name: init-dir
         image: busybox:1.28
diff --git a/build/installer/deploy/nvidia-device-plugin.yml b/build/installer/deploy/nvidia-device-plugin.yml
index 49adcf45..85ab63dd 100644
--- a/build/installer/deploy/nvidia-device-plugin.yml
+++ b/build/installer/deploy/nvidia-device-plugin.yml
@@ -44,6 +44,8 @@ spec:
       # be rescheduled after a failure.
       # See https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/
       priorityClassName: "system-node-critical"
+      nodeSelector:
+        gpu.bytetrade.io/cuda-supported: 'true'
       containers:
       - image: nvcr.io/nvidia/k8s-device-plugin:v0.16.1
         name: nvidia-device-plugin-ctr
diff --git a/build/installer/deploy/scheduler.yaml b/build/installer/deploy/scheduler.yaml
index 33fbcaff..c42043e2 100644
--- a/build/installer/deploy/scheduler.yaml
+++ b/build/installer/deploy/scheduler.yaml
@@ -28,6 +28,8 @@ spec:
     spec:
       runtimeClassName: nvidia # Explicitly request the runtime
       priorityClassName: system-node-critical
+      nodeSelector:
+        gpu.bytetrade.io/cuda-supported: 'true'
       initContainers:
       - name: init-dir
         image: busybox:1.28
diff --git a/build/installer/install.ps1 b/build/installer/install.ps1
index c23350ea..83a7c55c 100644
--- a/build/installer/install.ps1
+++ b/build/installer/install.ps1
@@ -38,7 +38,7 @@ if (-Not (Test-Path $CLI_PROGRAM_PATH)) {
   New-Item -Path $CLI_PROGRAM_PATH -ItemType Directory
 }
 
-$CLI_VERSION = "0.1.87"
+$CLI_VERSION = "0.1.89"
 $CLI_FILE = "olares-cli-v{0}_windows_{1}.tar.gz" -f $CLI_VERSION, $arch
 $CLI_URL = "{0}/{1}" -f $downloadUrl, $CLI_FILE
 $CLI_PATH = "{0}{1}" -f $CLI_PROGRAM_PATH, $CLI_FILE
diff --git a/build/installer/install.sh b/build/installer/install.sh
index dcff2fde..efe18797 100644
--- a/build/installer/install.sh
+++ b/build/installer/install.sh
@@ -74,7 +74,7 @@ if [ -z ${cdn_url} ]; then
     cdn_url="https://dc3p1870nn3cj.cloudfront.net"
 fi
 
-CLI_VERSION="0.1.87"
+CLI_VERSION="0.1.89"
 CLI_FILE="olares-cli-v${CLI_VERSION}_linux_${ARCH}.tar.gz"
 if [[ x"$os_type" == x"Darwin" ]]; then
     CLI_FILE="olares-cli-v${CLI_VERSION}_darwin_${ARCH}.tar.gz"