|
| 1 | +/** |
| 2 | +# Copyright (c) NVIDIA CORPORATION. All rights reserved. |
| 3 | +# |
| 4 | +# Licensed under the Apache License, Version 2.0 (the "License"); |
| 5 | +# you may not use this file except in compliance with the License. |
| 6 | +# You may obtain a copy of the License at |
| 7 | +# |
| 8 | +# http://www.apache.org/licenses/LICENSE-2.0 |
| 9 | +# |
| 10 | +# Unless required by applicable law or agreed to in writing, software |
| 11 | +# distributed under the License is distributed on an "AS IS" BASIS, |
| 12 | +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 13 | +# See the License for the specific language governing permissions and |
| 14 | +# limitations under the License. |
| 15 | +**/ |
| 16 | + |
| 17 | +package controllers |
| 18 | + |
| 19 | +import ( |
| 20 | + "bytes" |
| 21 | + "context" |
| 22 | + "errors" |
| 23 | + "testing" |
| 24 | + |
| 25 | + "github.com/go-logr/logr" |
| 26 | + "github.com/go-logr/zapr" |
| 27 | + "github.com/stretchr/testify/require" |
| 28 | + "go.uber.org/zap" |
| 29 | + "go.uber.org/zap/zapcore" |
| 30 | + metav1 "k8s.io/apimachinery/pkg/apis/meta/v1" |
| 31 | + "k8s.io/apimachinery/pkg/runtime" |
| 32 | + "k8s.io/apimachinery/pkg/types" |
| 33 | + "k8s.io/utils/ptr" |
| 34 | + ctrl "sigs.k8s.io/controller-runtime" |
| 35 | + "sigs.k8s.io/controller-runtime/pkg/client/fake" |
| 36 | + |
| 37 | + gpuv1 "github.com/NVIDIA/gpu-operator/api/nvidia/v1" |
| 38 | + nvidiav1alpha1 "github.com/NVIDIA/gpu-operator/api/nvidia/v1alpha1" |
| 39 | + "github.com/NVIDIA/gpu-operator/internal/validator" |
| 40 | +) |
| 41 | + |
| 42 | +// FakeConditionUpdater implements conditions.Updater |
| 43 | +// It always returns CustomError if set |
| 44 | +type FakeConditionUpdater struct { |
| 45 | + CustomError error |
| 46 | +} |
| 47 | + |
| 48 | +// SetConditionsError always returns CustomError if set |
| 49 | +func (f *FakeConditionUpdater) SetConditionsError(ctx context.Context, obj any, condType, msg string) error { |
| 50 | + return f.CustomError |
| 51 | +} |
| 52 | + |
| 53 | +// SetConditionsReady always returns CustomError if set |
| 54 | +func (f *FakeConditionUpdater) SetConditionsReady(ctx context.Context, obj any, condType, msg string) error { |
| 55 | + return f.CustomError |
| 56 | +} |
| 57 | + |
| 58 | +// FakeNodeSelectorValidator always returns CustomError if set |
| 59 | +type FakeNodeSelectorValidator struct { |
| 60 | + CustomError error |
| 61 | +} |
| 62 | + |
| 63 | +// Validate always returns CustomError if set |
| 64 | +func (f *FakeNodeSelectorValidator) Validate(ctx context.Context, cr *nvidiav1alpha1.NVIDIADriver) error { |
| 65 | + return f.CustomError |
| 66 | +} |
| 67 | + |
| 68 | +// newTestLogger creates a zap.Logger that writes to an in-memory buffer for testing |
| 69 | +func newTestLogger() (logr.Logger, *bytes.Buffer) { |
| 70 | + buf := &bytes.Buffer{} |
| 71 | + |
| 72 | + encoder := zapcore.NewConsoleEncoder(zapcore.EncoderConfig{ |
| 73 | + MessageKey: "msg", |
| 74 | + LevelKey: "level", |
| 75 | + NameKey: "logger", |
| 76 | + CallerKey: "caller", |
| 77 | + EncodeLevel: zapcore.CapitalLevelEncoder, |
| 78 | + EncodeCaller: zapcore.ShortCallerEncoder, |
| 79 | + EncodeName: zapcore.FullNameEncoder, |
| 80 | + }) |
| 81 | + |
| 82 | + core := zapcore.NewCore(encoder, zapcore.AddSync(buf), zapcore.DebugLevel) |
| 83 | + zapLogger := zap.New(core) |
| 84 | + |
| 85 | + return zapr.NewLogger(zapLogger), buf |
| 86 | +} |
| 87 | + |
| 88 | +// TestReconcile tests that reconciliation proceeds or skips based on the |
| 89 | +// ClusterPolicy and NVIDIADriver. Since Reconcile() does in-memory updates, |
| 90 | +// we just ensure it does not error out or the error is handled gracefully. |
| 91 | +func TestReconcile(t *testing.T) { |
| 92 | + scheme := runtime.NewScheme() |
| 93 | + require.NoError(t, nvidiav1alpha1.AddToScheme(scheme)) |
| 94 | + require.NoError(t, gpuv1.AddToScheme(scheme)) |
| 95 | + |
| 96 | + tests := []struct { |
| 97 | + name string |
| 98 | + useCRD *bool |
| 99 | + spec nvidiav1alpha1.NVIDIADriverSpec |
| 100 | + validator validator.Validator |
| 101 | + error error |
| 102 | + expectedLog string |
| 103 | + }{ |
| 104 | + { |
| 105 | + name: "ClusterPolicy has driver CRD false → reconciliation skips driver", |
| 106 | + useCRD: ptr.To(false), |
| 107 | + validator: &FakeNodeSelectorValidator{ |
| 108 | + CustomError: errors.New("fake list error"), |
| 109 | + }, |
| 110 | + error: nil, |
| 111 | + expectedLog: "useNvidiaDriverCRD is not enabled in ClusterPolicy", |
| 112 | + }, |
| 113 | + { |
| 114 | + name: "ClusterPolicy has driver CRD true but validator errors", |
| 115 | + useCRD: ptr.To(true), |
| 116 | + validator: &FakeNodeSelectorValidator{ |
| 117 | + CustomError: errors.New("fake list error"), |
| 118 | + }, |
| 119 | + error: nil, |
| 120 | + expectedLog: "nodeSelector validation failed", |
| 121 | + }, |
| 122 | + { |
| 123 | + name: "driver CRD true, no validator errors, use precompiled drivers and GDS enabled", |
| 124 | + useCRD: ptr.To(true), |
| 125 | + spec: nvidiav1alpha1.NVIDIADriverSpec{ |
| 126 | + UsePrecompiled: ptr.To(true), |
| 127 | + GPUDirectStorage: &nvidiav1alpha1.GPUDirectStorageSpec{ |
| 128 | + Enabled: ptr.To(true), |
| 129 | + }, |
| 130 | + }, |
| 131 | + validator: &FakeNodeSelectorValidator{ |
| 132 | + CustomError: nil, |
| 133 | + }, |
| 134 | + error: nil, |
| 135 | + expectedLog: "GPUDirect Storage driver (nvidia-fs) and/or GDRCopy driver is not supported along with pre-compiled NVIDIA drivers", |
| 136 | + }, |
| 137 | + } |
| 138 | + |
| 139 | + for _, tc := range tests { |
| 140 | + t.Run(tc.name, func(t *testing.T) { |
| 141 | + logger, buf := newTestLogger() |
| 142 | + ctx := ctrl.LoggerInto(context.Background(), logger) |
| 143 | + driver := &nvidiav1alpha1.NVIDIADriver{ |
| 144 | + ObjectMeta: metav1.ObjectMeta{ |
| 145 | + Name: "test-driver", |
| 146 | + Namespace: "default", |
| 147 | + }, |
| 148 | + Spec: tc.spec, |
| 149 | + } |
| 150 | + |
| 151 | + cp := &gpuv1.ClusterPolicy{ |
| 152 | + ObjectMeta: metav1.ObjectMeta{Name: "default"}, |
| 153 | + Spec: gpuv1.ClusterPolicySpec{ |
| 154 | + Driver: gpuv1.DriverSpec{ |
| 155 | + UseNvidiaDriverCRD: tc.useCRD, |
| 156 | + }, |
| 157 | + }, |
| 158 | + } |
| 159 | + |
| 160 | + // Initialize fake client with ClusterPolicy (driver optional) |
| 161 | + clientBuilder := fake.NewClientBuilder().WithScheme(scheme).WithObjects(cp, driver) |
| 162 | + client := clientBuilder.Build() |
| 163 | + |
| 164 | + updater := &FakeConditionUpdater{} |
| 165 | + |
| 166 | + reconciler := &NVIDIADriverReconciler{ |
| 167 | + Client: client, |
| 168 | + Scheme: scheme, |
| 169 | + conditionUpdater: updater, |
| 170 | + nodeSelectorValidator: tc.validator, |
| 171 | + } |
| 172 | + |
| 173 | + req := ctrl.Request{ |
| 174 | + NamespacedName: types.NamespacedName{ |
| 175 | + Name: driver.Name, |
| 176 | + Namespace: driver.Namespace, |
| 177 | + }, |
| 178 | + } |
| 179 | + |
| 180 | + _, err := reconciler.Reconcile(ctx, req) |
| 181 | + |
| 182 | + if tc.error != nil { |
| 183 | + require.Error(t, err) |
| 184 | + require.EqualError(t, err, tc.error.Error()) |
| 185 | + } else { |
| 186 | + require.NoError(t, err) |
| 187 | + } |
| 188 | + |
| 189 | + logs := buf.String() |
| 190 | + if tc.expectedLog != "" { |
| 191 | + require.Contains(t, logs, tc.expectedLog) |
| 192 | + } |
| 193 | + }) |
| 194 | + } |
| 195 | +} |
0 commit comments