Merge pull request #4817 from Azure/gather-observability-query-embedding

gather-observability: embed query config into binary

Author: openshift-merge-bot[bot]

test/E2ELocal.mk

@@ -47,8 +47,7 @@ e2e-local/gather-observability: $(ARO_HCP_TESTS) $(TEMPLATIZE)
 		--subscription-id "$$(az account show --query id -o tsv)" \
 		--output $(OBSERVABILITY_OUTPUT) \
 		--severity-threshold Sev3 \
-		--start-time-fallback "$$(date -u -v-1H +%Y-%m-%dT%H:%M:%SZ)" \
-		--queries-config $(DIR)/cmd/aro-hcp-tests/gather-observability/queries.yaml
+		--start-time-fallback "$$(date -u -v-1H +%Y-%m-%dT%H:%M:%SZ)"
 	@echo "Observability artifacts written to $(OBSERVABILITY_OUTPUT)"
 .PHONY: e2e-local/gather-observability
 

test/cmd/aro-hcp-tests/gather-observability/chart_test.go

@@ -17,8 +17,6 @@ package gatherobservability
 import (
 	"encoding/json"
 	"math"
-	"os"
-	"path/filepath"
 	"strings"
 	"testing"
 )
@@ -497,13 +495,7 @@ func TestLoadQueriesConfig(t *testing.T) {
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
 			t.Parallel()
-			tmpDir := t.TempDir()
-			cfgPath := filepath.Join(tmpDir, "queries.yaml")
-			if err := os.WriteFile(cfgPath, []byte(tt.yaml), 0644); err != nil {
-				t.Fatalf("failed to write temp config: %v", err)
-			}
-
-			cfg, err := loadQueriesConfig(cfgPath)
+			cfg, err := parseQueriesConfig([]byte(tt.yaml))
 			if tt.wantErr != "" {
 				if err == nil {
 					t.Fatalf("expected error containing %q, got nil", tt.wantErr)
@@ -523,10 +515,13 @@ func TestLoadQueriesConfig(t *testing.T) {
 	}
 }
 
-func TestLoadQueriesConfigFileNotFound(t *testing.T) {
+func TestLoadQueriesConfigEmbedded(t *testing.T) {
 	t.Parallel()
-	_, err := loadQueriesConfig("/nonexistent/path/queries.yaml")
-	if err == nil {
-		t.Fatal("expected error for missing file, got nil")
+	cfg, err := loadQueriesConfig()
+	if err != nil {
+		t.Fatalf("embedded queries.yaml should parse without error: %v", err)
+	}
+	if len(cfg.Queries) == 0 {
+		t.Fatal("embedded queries.yaml should contain at least one query")
 	}
 }

test/cmd/aro-hcp-tests/gather-observability/options.go

@@ -62,7 +62,6 @@ func BindOptions(opts *RawOptions, cmd *cobra.Command) error {
 	cmd.Flags().StringVar(&opts.RenderedConfig, "rendered-config", opts.RenderedConfig, "Path to the rendered configuration YAML file.")
 	cmd.Flags().StringVar(&opts.SubscriptionID, "subscription-id", opts.SubscriptionID, "Azure subscription ID.")
 	cmd.Flags().StringVar(&opts.StartTimeFallback, "start-time-fallback", opts.StartTimeFallback, "Optional RFC3339 time to use as start time fallback when steps and test timing are unavailable.")
-	cmd.Flags().StringVar(&opts.QueriesConfigPath, "queries-config", opts.QueriesConfigPath, "Path to YAML file specifying PromQL queries to execute and chart.")
 	cmd.Flags().StringVar(&opts.SeverityThreshold, "severity-threshold", opts.SeverityThreshold, "Include alerts at this severity level or more critical (Sev0=critical .. Sev4=verbose). E.g. Sev2 includes Sev0, Sev1, Sev2. If not set, all severities are shown.")
 	return nil
 }
@@ -73,7 +72,6 @@ type RawOptions struct {
 	RenderedConfig    string
 	SubscriptionID    string
 	StartTimeFallback string
-	QueriesConfigPath string
 	SeverityThreshold string
 }
 
@@ -199,27 +197,25 @@ func (o *ValidatedOptions) Complete(ctx context.Context) (*Options, error) {
 		cred:              cred,
 	}
 
-	if o.QueriesConfigPath != "" {
-		queries, err := loadQueriesConfig(o.QueriesConfigPath)
-		if err != nil {
-			return nil, fmt.Errorf("failed to load queries config: %w", err)
-		}
-		completed.Queries = queries
-		logger.Info("loaded queries config", "path", o.QueriesConfigPath, "queries", len(queries.Queries))
+	queries, err := loadQueriesConfig()
+	if err != nil {
+		return nil, fmt.Errorf("failed to load queries config: %w", err)
+	}
+	completed.Queries = queries
+	logger.Info("loaded embedded queries config", "queries", len(queries.Queries))
 
-		// Resolve Prometheus endpoints eagerly so we fail fast
-		completed.SvcPromEndpoint, err = lookupPrometheusEndpoint(ctx, cred, o.SubscriptionID, regionRG, svcWorkspace)
-		if err != nil {
-			return nil, fmt.Errorf("failed to look up svc Prometheus endpoint for workspace %s in %s: %w", svcWorkspace, regionRG, err)
-		}
-		logger.Info("resolved svc Prometheus endpoint", "endpoint", completed.SvcPromEndpoint)
+	// Resolve Prometheus endpoints eagerly so we fail fast
+	completed.SvcPromEndpoint, err = lookupPrometheusEndpoint(ctx, cred, o.SubscriptionID, regionRG, svcWorkspace)
+	if err != nil {
+		return nil, fmt.Errorf("failed to look up svc Prometheus endpoint for workspace %s in %s: %w", svcWorkspace, regionRG, err)
+	}
+	logger.Info("resolved svc Prometheus endpoint", "endpoint", completed.SvcPromEndpoint)
 
-		completed.HcpPromEndpoint, err = lookupPrometheusEndpoint(ctx, cred, o.SubscriptionID, regionRG, hcpWorkspace)
-		if err != nil {
-			return nil, fmt.Errorf("failed to look up hcp Prometheus endpoint for workspace %s in %s: %w", hcpWorkspace, regionRG, err)
-		}
-		logger.Info("resolved hcp Prometheus endpoint", "endpoint", completed.HcpPromEndpoint)
+	completed.HcpPromEndpoint, err = lookupPrometheusEndpoint(ctx, cred, o.SubscriptionID, regionRG, hcpWorkspace)
+	if err != nil {
+		return nil, fmt.Errorf("failed to look up hcp Prometheus endpoint for workspace %s in %s: %w", hcpWorkspace, regionRG, err)
 	}
+	logger.Info("resolved hcp Prometheus endpoint", "endpoint", completed.HcpPromEndpoint)
 
 	return &Options{completedOptions: completed}, nil
 }

test/cmd/aro-hcp-tests/gather-observability/promql.go

@@ -22,11 +22,12 @@ import (
 	"io"
 	"net/http"
 	"net/url"
-	"os"
 	"strconv"
 	"strings"
 	"time"
 
+	_ "embed"
+
 	"sigs.k8s.io/yaml"
 
 	"github.com/Azure/azure-sdk-for-go/sdk/azcore"
@@ -68,11 +69,14 @@ type PrometheusResult struct {
 	Values [][]any           `json:"values"` // each element is [unix_timestamp_float, "string_value"]
 }
 
-func loadQueriesConfig(path string) (*QueriesConfig, error) {
-	data, err := os.ReadFile(path)
-	if err != nil {
-		return nil, fmt.Errorf("failed to read queries config %s: %w", path, err)
-	}
+//go:embed queries.yaml
+var defaultQueriesYAML []byte
+
+func loadQueriesConfig() (*QueriesConfig, error) {
+	return parseQueriesConfig(defaultQueriesYAML)
+}
+
+func parseQueriesConfig(data []byte) (*QueriesConfig, error) {
 	var cfg QueriesConfig
 	if err := yaml.Unmarshal(data, &cfg); err != nil {
 		return nil, fmt.Errorf("failed to parse queries config: %w", err)