ITWorx-Pulse-Public/internal/alertcontrol/repository_test.go

package alertcontrol

import (
	"context"
	"os"
	"sync"
	"testing"
	"time"

	"github.com/itworx/pulse/internal/database"
)

func TestPostgreSQLControlsAreExpiringAuditedAndIdempotent(t *testing.T) {
	dsn := os.Getenv("PULSE_TEST_DATABASE_URL")
	if dsn == "" {
		t.Skip("PULSE_TEST_DATABASE_URL is not set")
	}
	ctx, cancel := context.WithTimeout(context.Background(), 45*time.Second)
	defer cancel()
	pool, err := database.NewPool(ctx, database.Config{URL: dsn, MaxConns: 8, MinConns: 1})
	if err != nil {
		t.Fatal(err)
	}
	defer pool.Close()
	if err := database.Migrate(ctx, pool); err != nil {
		t.Fatal(err)
	}
	repo := Repository{Pool: pool}
	now := time.Now().UTC().Truncate(time.Microsecond)
	silenceID, maintenanceID, expiringID := NewID(), NewID(), NewID()
	t.Cleanup(func() {
		cleanupCtx, cleanupCancel := context.WithTimeout(context.Background(), 10*time.Second)
		defer cleanupCancel()
		_, _ = pool.Exec(cleanupCtx, `DELETE FROM alert_silences WHERE id = ANY($1::uuid[])`, []string{silenceID, expiringID})
		_, _ = pool.Exec(cleanupCtx, `DELETE FROM maintenance_windows WHERE id = $1::uuid`, maintenanceID)
	})

	silence, err := repo.CreateSilence(ctx, "operator-1", Silence{ID: silenceID, Name: "planned deploy", Reason: "change window", Owner: "operator-1", Matchers: Matcher{Severities: []string{"critical"}}, StartsAt: now.Add(-time.Minute), ExpiresAt: now.Add(time.Hour)})
	if err != nil {
		t.Fatal(err)
	}
	if silence.State != StateActive || silence.Revision != 1 {
		t.Fatalf("unexpected silence: %#v", silence)
	}
	window, err := repo.CreateMaintenance(ctx, "operator-1", MaintenanceWindow{ID: maintenanceID, Name: "maintenance", Reason: "firmware", Selector: Matcher{EntityTypes: []string{"host"}}, StartsAt: now.Add(-time.Minute), EndsAt: now.Add(time.Hour)})
	if err != nil {
		t.Fatal(err)
	}
	if window.State != StateActive {
		t.Fatalf("maintenance state = %s", window.State)
	}
	items, err := repo.ListSilences(ctx, 100, now)
	if err != nil || len(items) != 1 {
		t.Fatalf("list silences: %d, %v", len(items), err)
	}

	expiring, err := repo.CreateSilence(ctx, "operator-1", Silence{ID: expiringID, Name: "short", Reason: "test expiry", Owner: "operator-1", Matchers: Matcher{}, StartsAt: now.Add(-time.Minute), ExpiresAt: now.Add(-time.Second)})
	if err != nil {
		t.Fatal(err)
	}
	if expiring.State != StateExpired {
		t.Fatalf("expired control before job = %s", expiring.State)
	}
	first, err := repo.Expire(ctx, now)
	if err != nil {
		t.Fatal(err)
	}
	second, err := repo.Expire(ctx, now)
	if err != nil {
		t.Fatal(err)
	}
	if first.Silences != 1 || first.MaintenanceWindows != 0 || second != (ExpiryResult{}) {
		t.Fatalf("expiry not idempotent: first=%#v second=%#v", first, second)
	}

	concurrentID := NewID()
	defer func() {
		_, _ = pool.Exec(context.Background(), `DELETE FROM alert_silences WHERE id=$1::uuid`, concurrentID)
	}()
	if _, err := repo.CreateSilence(ctx, "operator-1", Silence{ID: concurrentID, Name: "concurrent", Reason: "test", Owner: "operator-1", Matchers: Matcher{}, StartsAt: now.Add(-time.Minute), ExpiresAt: now.Add(-time.Second)}); err != nil {
		t.Fatal(err)
	}
	var wg sync.WaitGroup
	results := make(chan ExpiryResult, 2)
	errorsCh := make(chan error, 2)
	for i := 0; i < 2; i++ {
		wg.Add(1)
		go func() {
			defer wg.Done()
			result, err := repo.Expire(ctx, now.Add(time.Second))
			if err != nil {
				errorsCh <- err
				return
			}
			results <- result
		}()
	}
	wg.Wait()
	close(results)
	close(errorsCh)
	total := 0
	for result := range results {
		total += result.Silences + result.MaintenanceWindows
	}
	for err := range errorsCh {
		t.Fatal(err)
	}
	if total != 1 {
		t.Fatalf("concurrent expiry count = %d, want 1", total)
	}

	if _, err := repo.CreateSilence(ctx, "operator-1", Silence{ID: silenceID, Name: "duplicate", Reason: "duplicate", Owner: "operator-1", Matchers: Matcher{}, StartsAt: now, ExpiresAt: now.Add(time.Hour)}); err == nil {
		t.Fatal("duplicate silence should fail")
	}
	if _, err := repo.RevokeSilence(ctx, silenceID, "operator-1", silence.Revision, now); err != nil {
		t.Fatal(err)
	}
	if _, err := repo.RevokeSilence(ctx, silenceID, "operator-1", silence.Revision+1, now); err == nil {
		t.Fatal("repeated revoke should conflict")
	}
}