crt: onboard nomad-device-nvidia to crt

.github/CODEOWNERS

@@ -0,0 +1,5 @@
+* @hashicorp/nomad-eng
+
+# release configuration
+/.release/                              @hashicorp/release-engineering
+/.github/workflows/build.yml            @hashicorp/release-engineering

.github/workflows/build.yml

@@ -0,0 +1,89 @@
+name: build
+
+on:
+  push:
+  workflow_dispatch:
+
+env:
+  PKG_NAME: "nomad-device-nvidia"
+
+jobs:
+  get-go-version:
+    name: "Determine Go toolchain version"
+    runs-on: ubuntu-20.04
+    outputs:
+      go-version: ${{ steps.get-go-version.outputs.go-version }}
+    steps:
+      - uses: actions/checkout@v2
+      - name: Determine Go version
+        id: get-go-version
+        run: |
+          echo "Building with Go $(cat .go-version)"
+          echo "::set-output name=go-version::$(cat .go-version)"
+
+  get-product-version:
+    runs-on: ubuntu-20.04
+    outputs:
+      product-version: ${{ steps.get-product-version.outputs.product-version }}
+    steps:
+      - uses: actions/checkout@v2
+      - name: get product version
+        id: get-product-version
+        run: |
+          make version
+          echo "::set-output name=product-version::$(make version)"
+
+  generate-metadata-file:
+    needs: get-product-version
+    runs-on: ubuntu-20.04
+    outputs:
+      filepath: ${{ steps.generate-metadata-file.outputs.filepath }}
+    steps:
+      - name: "Checkout directory"
+        uses: actions/checkout@v2
+      - name: Generate metadata file
+        id: generate-metadata-file
+        uses: hashicorp/actions-generate-metadata@v1
+        with:
+          version: ${{ needs.get-product-version.outputs.product-version }}
+          product: ${{ env.PKG_NAME }}
+          repositoryOwner: "hashicorp"
+      - uses: actions/upload-artifact@v2
+        if: ${{ !env.ACT }}
+        with:
+          name: metadata.json
+          path: ${{ steps.generate-metadata-file.outputs.filepath }}
+
+  build-linux:
+    needs:
+      - get-go-version
+      - get-product-version
+    runs-on: ubuntu-20.04
+    strategy:
+      matrix:
+        goos: ["linux"]
+        goarch: ["amd64"]
+      fail-fast: true
+
+    name: Go ${{ needs.get-go-version.outputs.go-version }} ${{ matrix.goos }} ${{ matrix.goarch }} build
+
+    steps:
+      - uses: actions/checkout@v2
+      - name: Setup go
+        uses: actions/setup-go@v2
+        with:
+          go-version: ${{ needs.get-go-version.outputs.go-version }}
+      - name: Build
+        env:
+          GOOS: ${{ matrix.goos }}
+          GOARCH: ${{ matrix.goarch }}
+        run: |
+          make pkg/${{ matrix.goos }}_${{ matrix.goarch }}.zip
+          mv \
+            pkg/${{ matrix.goos }}_${{ matrix.goarch }}.zip \
+            ${{ env.PKG_NAME }}_${{ needs.get-product-version.outputs.product-version }}_${{ matrix.goos }}_${{ matrix.goarch }}.zip
+      - uses: actions/upload-artifact@v2
+        if: ${{ !env.ACT }}
+        with:
+          name: ${{ env.PKG_NAME }}_${{ needs.get-product-version.outputs.product-version }}_${{ matrix.goos }}_${{ matrix.goarch }}.zip
+          path: ${{ env.PKG_NAME }}_${{ needs.get-product-version.outputs.product-version }}_${{ matrix.goos }}_${{ matrix.goarch }}.zip

.github/workflows/tests.yaml

@@ -0,0 +1,16 @@
+name: Run CI Tests
+on: [push]
+env:
+  GOBIN: /opt/bin
+  GO_VERSION: 1.17.9
+jobs:
+  run-tests:
+    runs-on: ubuntu-20.04
+    steps:
+      - uses: actions/checkout@v2
+      - uses: magnetikonline/action-golang-cache@v1
+        with:
+          go-version: ${{env.GO_VERSION}}
+      - name: Run Go Tests
+        run: |
+          make test

.gitignore

@@ -17,3 +17,4 @@
 # Build output
 bin/
 pkg/
+nomad-device-nvidia

.go-version

@@ -0,0 +1 @@
+1.17.9

.release/ci.hcl

@@ -0,0 +1,132 @@
+schema = "1"
+
+project "nomad-device-nvidia" {
+  team = "nomad"
+  slack {
+    notification_channel = "C03B5EWFW01"
+  }
+  github {
+    organization = "hashicorp"
+    repository   = "nomad-device-nvidia"
+    release_branches = [
+      "main",
+      "crt-onboard",
+    ]
+  }
+}
+
+event "merge" {
+  // "entrypoint" to use if build is not run automatically
+  // i.e. send "merge" complete signal to orchestrator to trigger build
+}
+
+event "build" {
+  depends = ["merge"]
+  action "build" {
+    organization = "hashicorp"
+    repository   = "nomad-device-nvidia"
+    workflow     = "build"
+  }
+}
+
+event "upload-dev" {
+  depends = ["build"]
+  action "upload-dev" {
+    organization = "hashicorp"
+    repository   = "crt-workflows-common"
+    workflow     = "upload-dev"
+    depends      = ["build"]
+  }
+
+  notification {
+    on = "fail"
+  }
+}
+
+event "security-scan-binaries" {
+  depends = ["upload-dev"]
+  action "security-scan-binaries" {
+    organization = "hashicorp"
+    repository   = "crt-workflows-common"
+    workflow     = "security-scan-binaries"
+    config       = "security-scan.hcl"
+  }
+
+  notification {
+    on = "fail"
+  }
+}
+
+event "sign" {
+  depends = ["security-scan-binaries"]
+  action "sign" {
+    organization = "hashicorp"
+    repository   = "crt-workflows-common"
+    workflow     = "sign"
+  }
+
+  notification {
+    on = "fail"
+  }
+}
+
+event "verify" {
+  depends = ["sign"]
+  action "verify" {
+    organization = "hashicorp"
+    repository   = "crt-workflows-common"
+    workflow     = "verify"
+  }
+
+  notification {
+    on = "always"
+  }
+}
+
+event "fossa-scan" {
+  depends = ["verify"]
+  action "fossa-scan" {
+    organization = "hashicorp"
+    repository   = "crt-workflows-common"
+    workflow     = "fossa-scan"
+  }
+}
+
+## These are promotion and post-publish events
+## they should be added to the end of the file after the verify event stanza.
+
+event "trigger-staging" {
+  // This event is dispatched by the bob trigger-promotion command
+  // and is required - do not delete.
+}
+
+event "promote-staging" {
+  depends = ["trigger-staging"]
+  action "promote-staging" {
+    organization = "hashicorp"
+    repository   = "crt-workflows-common"
+    workflow     = "promote-staging"
+  }
+
+  notification {
+    on = "always"
+  }
+}
+
+event "trigger-production" {
+  // This event is dispatched by the bob trigger-promotion command
+  // and is required - do not delete.
+}
+
+event "promote-production" {
+  depends = ["trigger-production"]
+  action "promote-production" {
+    organization = "hashicorp"
+    repository   = "crt-workflows-common"
+    workflow     = "promote-production"
+  }
+
+  notification {
+    on = "always"
+  }
+}

.release/security-scan.hcl

@@ -0,0 +1,8 @@
+binary {
+  secrets    = true
+  go_modules = true
+  #TODO: enable OSV scan once dependencies are updated.
+  osv       = false
+  oss_index = false
+  nvd       = false
+}

GNUmakefile

@@ -1,13 +1,55 @@
-default: build
+SHELL = bash
+default: help
 
-build: bin/nomad-device-nvidia
+GIT_COMMIT := $(shell git rev-parse --short HEAD)
+GIT_DIRTY := $(if $(shell git status --porcelain),+CHANGES)
+
+GO_LDFLAGS := "-X github.com/hashicorp/nomad-autoscaler/version.GitCommit=$(GIT_COMMIT)$(GIT_DIRTY)"
+
+HELP_FORMAT="    \033[36m%-25s\033[0m %s\n"
+.PHONY: help
+help: ## Display this usage information
+	@echo "Valid targets:"
+	@grep -E '^[^ ]+:.*?## .*$$' $(MAKEFILE_LIST) | \
+		sort | \
+		awk 'BEGIN {FS = ":.*?## "}; \
+			{printf $(HELP_FORMAT), $$1, $$2}'
+	@echo ""
+
+.PHONY: clean
+clean: ## Cleanup previous build
+	@echo "==> Cleanup previous build"
+	rm -f ./bin/nomad-device-nvidia
+
+pkg/%/nomad-device-nvidia: GO_OUT ?= $@
+pkg/%/nomad-device-nvidia: ## Build nomad-device-nvidia for GOOS_GOARCH, e.g. pkg/linux_amd64/nomad-device-nvidia
+	@echo "==> Building $@ with tags $(GO_TAGS)..."
+	@GOOS=$(firstword $(subst _, ,$*)) \
+		GOARCH=$(lastword $(subst _, ,$*)) \
+		go build -trimpath -ldflags $(GO_LDFLAGS) -tags "$(GO_TAGS)" -o $(GO_OUT) cmd/main.go
+
+.PRECIOUS: pkg/%/nomad-device-nvidia
+pkg/%.zip: pkg/%/nomad-device-nvidia ## Build and zip nomad-device-nvidia for GOOS_GOARCH, e.g. pkg/linux_amd64.zip
+	@echo "==> Packaging for $@..."
+	zip -j $@ $(dir $<)*
+
+.PHONY: dev
+dev: clean bin/nomad-device-nvidia ## Build the nomad-device-nvidia plugin
 
 bin/nomad-device-nvidia:
+	@echo "==> Building device driver ..."
 	mkdir -p bin
-	go build \
-		-o ./bin/nomad-device-nvidia \
-		./cmd/main.go
+	go build -o bin/nomad-device-nvidia cmd/main.go
 
 .PHONY: test
-test:
-	go test -v ./...
+test: ## Run unit tests
+	@echo "==> Running tests ..."
+	go test -v -race ./...
+
+.PHONY: version
+version: ## Get the current version string
+ifneq (,$(wildcard version/version_ent.go))
+	@$(CURDIR)/scripts/version.sh version/version.go version/version_ent.go
+else
+	@$(CURDIR)/scripts/version.sh version/version.go version/version.go
+endif

README.md

@@ -4,8 +4,6 @@ This repository provides an implementation of a Nomad
 [`device`](https://www.nomadproject.io/docs/job-specification/device) plugin
 for Nvidia GPUs.
 
-**Note: this package is currently embedded in Nomad itself but the Nomad team intends to externalize the driver.**
-
 ## Behavior
 
 The Nvidia device plugin uses

cmd/main.go

@@ -3,9 +3,8 @@ package main
 import (
 	"context"
 
-	log "github.com/hashicorp/go-hclog"
-
-	"github.com/hashicorp/nomad/devices/gpu/nvidia"
+	"github.com/hashicorp/go-hclog"
+	"github.com/hashicorp/nomad-device-nvidia"
 	"github.com/hashicorp/nomad/plugins"
 )
 
@@ -15,6 +14,6 @@ func main() {
 }
 
 // factory returns a new instance of the Nvidia GPU plugin
-func factory(ctx context.Context, log log.Logger) interface{} {
+func factory(ctx context.Context, log hclog.Logger) interface{} {
 	return nvidia.NewNvidiaDevice(ctx, log)
 }

device.go

@@ -1,4 +1,4 @@
-package main
+package nvidia
 
 import (
 	"context"
@@ -7,9 +7,9 @@ import (
 	"sync"
 	"time"
 
-	log "github.com/hashicorp/go-hclog"
+	"github.com/hashicorp/go-hclog"
+	"github.com/hashicorp/nomad-device-nvidia/nvml"
 	"github.com/hashicorp/nomad-device-nvidia/version"
-	"github.com/hashicorp/nomad/devices/gpu/nvidia/nvml"
 	"github.com/hashicorp/nomad/helper/pluginutils/loader"
 	"github.com/hashicorp/nomad/plugins/base"
 	"github.com/hashicorp/nomad/plugins/device"
@@ -45,7 +45,7 @@ var (
 	// PluginConfig is the nvidia factory function registered in the
 	// plugin catalog.
 	PluginConfig = &loader.InternalPluginConfig{
-		Factory: func(ctx context.Context, l log.Logger) interface{} { return NewNvidiaDevice(ctx, l) },
+		Factory: func(ctx context.Context, l hclog.Logger) interface{} { return NewNvidiaDevice(ctx, l) },
 	}
 
 	// pluginInfo describes the plugin
@@ -102,11 +102,11 @@ type NvidiaDevice struct {
 	devices    map[string]struct{}
 	deviceLock sync.RWMutex
 
-	logger log.Logger
+	logger hclog.Logger
 }
 
 // NewNvidiaDevice returns a new nvidia device plugin.
-func NewNvidiaDevice(_ context.Context, log log.Logger) *NvidiaDevice {
+func NewNvidiaDevice(_ context.Context, log hclog.Logger) *NvidiaDevice {
 	nvmlClient, err := nvml.NewNvmlClient()
 	logger := log.Named(pluginName)
 	if err != nil && err.Error() != nvml.UnavailableLib.Error() {