fix(update): avoid OOM during large SQLite snapshot imports

CHANGELOG.md

@@ -4,6 +4,10 @@
 
 ### Changes
 
+### Fixes
+
+- Kept large Git snapshot imports and FTS rebuilds from exhausting memory on small hosts by using file-backed SQLite temp storage and a bounded import cache. (#65) Thanks @hxy91819.
+
 ## 0.7.2 - 2026-05-11
 
 ### Changes

internal/share/import_memory_test.go

@@ -0,0 +1,220 @@
+package share
+
+import (
+	"compress/gzip"
+	"context"
+	"encoding/json"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strconv"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/require"
+
+	"github.com/openclaw/discrawl/internal/store"
+)
+
+func TestImportRealSnapshot(t *testing.T) {
+	repo := strings.TrimSpace(os.Getenv("DISCRAWL_REAL_REPO"))
+	if repo == "" {
+		t.Skip("set DISCRAWL_REAL_REPO to run real snapshot import validation")
+	}
+
+	ctx := context.Background()
+	dst, err := store.Open(ctx, filepath.Join(t.TempDir(), "dst.db"))
+	require.NoError(t, err)
+	defer func() { _ = dst.Close() }()
+
+	_, changed, err := ImportIfChanged(ctx, dst, Options{
+		RepoPath: repo,
+		Branch:   "main",
+		Progress: func(p ImportProgress) {
+			if p.Phase == "start" || p.Phase == "rebuild_fts" || p.Phase == "done" {
+				t.Logf("import progress phase=%s total_rows=%d", p.Phase, p.TotalRows)
+			}
+		},
+	})
+	require.NoError(t, err)
+	require.True(t, changed)
+
+	var messageCount int
+	require.NoError(t, dst.DB().QueryRowContext(ctx, `select count(*) from messages`).Scan(&messageCount))
+	require.Positive(t, messageCount)
+	var ftsCount int
+	require.NoError(t, dst.DB().QueryRowContext(ctx, `select count(*) from message_fts`).Scan(&ftsCount))
+	require.Equal(t, messageCount, ftsCount)
+}
+
+func TestImportMemoryBounded(t *testing.T) {
+	if os.Getenv("DISCRAWL_OOM_REGRESSION") != "1" {
+		t.Skip("set DISCRAWL_OOM_REGRESSION=1 to run the memory-bounded import regression")
+	}
+
+	ctx := context.Background()
+	repo := t.TempDir()
+	messageRows := envInt(t, "DISCRAWL_OOM_ROWS", 80000)
+	textBytes := envInt(t, "DISCRAWL_OOM_TEXT_BYTES", 2048)
+	t.Logf("building synthetic snapshot rows=%d text_bytes=%d", messageRows, textBytes)
+	writeSyntheticMemorySnapshot(t, repo, messageRows, textBytes)
+	t.Log("synthetic snapshot built; starting import")
+
+	dst, err := store.Open(ctx, filepath.Join(t.TempDir(), "dst.db"))
+	require.NoError(t, err)
+	defer func() { _ = dst.Close() }()
+
+	var progress []ImportProgress
+	_, changed, err := ImportIfChanged(ctx, dst, Options{
+		RepoPath: repo,
+		Branch:   "main",
+		Progress: func(p ImportProgress) { progress = append(progress, p) },
+	})
+	require.NoError(t, err)
+	require.True(t, changed)
+	require.Contains(t, progressPhases(progress), "rebuild_fts")
+
+	needle := "oomunique000042"
+	results, err := dst.SearchMessages(ctx, store.SearchOptions{Query: needle, Limit: 10})
+	require.NoError(t, err)
+	require.Len(t, results, 1)
+	require.Contains(t, results[0].Content, needle)
+}
+
+func writeSyntheticMemorySnapshot(t *testing.T, repo string, messageRows, textBytes int) {
+	t.Helper()
+	generatedAt := time.Now().UTC()
+	updatedAt := generatedAt.Format(time.RFC3339Nano)
+	guildFile := "tables/guilds/000000.jsonl.gz"
+	channelFile := "tables/channels/000000.jsonl.gz"
+	memberFile := "tables/members/000000.jsonl.gz"
+	messageFile := "tables/messages/000000.jsonl.gz"
+
+	writeJSONLGzip(t, repo, guildFile, func(enc *json.Encoder) int {
+		require.NoError(t, enc.Encode(map[string]any{
+			"id":         "g1",
+			"name":       "Guild",
+			"icon":       "",
+			"raw_json":   `{}`,
+			"updated_at": updatedAt,
+		}))
+		return 1
+	})
+	writeJSONLGzip(t, repo, channelFile, func(enc *json.Encoder) int {
+		require.NoError(t, enc.Encode(map[string]any{
+			"id":                "c1",
+			"guild_id":          "g1",
+			"parent_id":         "",
+			"kind":              "text",
+			"name":              "general",
+			"topic":             "",
+			"position":          0,
+			"is_nsfw":           false,
+			"is_archived":       false,
+			"is_locked":         false,
+			"is_private_thread": false,
+			"thread_parent_id":  "",
+			"archive_timestamp": "",
+			"raw_json":          `{}`,
+			"updated_at":        updatedAt,
+		}))
+		return 1
+	})
+	writeJSONLGzip(t, repo, memberFile, func(enc *json.Encoder) int {
+		require.NoError(t, enc.Encode(map[string]any{
+			"guild_id":      "g1",
+			"user_id":       "u1",
+			"username":      "peter",
+			"global_name":   "",
+			"display_name":  "Peter",
+			"nick":          "",
+			"discriminator": "",
+			"avatar":        "",
+			"bot":           false,
+			"joined_at":     "",
+			"role_ids_json": `[]`,
+			"raw_json":      `{"bio":"memory regression profile"}`,
+			"updated_at":    updatedAt,
+		}))
+		return 1
+	})
+	writeJSONLGzip(t, repo, messageFile, func(enc *json.Encoder) int {
+		for i := range messageRows {
+			messageID := strconv.FormatInt(1456744319972282449+int64(i), 10)
+			unique := fmt.Sprintf("oomunique%06d", i)
+			content := syntheticMemoryContent(unique, textBytes)
+			require.NoError(t, enc.Encode(map[string]any{
+				"id":                  messageID,
+				"guild_id":            "g1",
+				"channel_id":          "c1",
+				"author_id":           "u1",
+				"message_type":        0,
+				"created_at":          updatedAt,
+				"edited_at":           "",
+				"deleted_at":          "",
+				"content":             content,
+				"normalized_content":  content,
+				"reply_to_message_id": "",
+				"pinned":              false,
+				"has_attachments":     false,
+				"raw_json":            `{"author":{"username":"Peter"}}`,
+				"updated_at":          updatedAt,
+			}))
+		}
+		return messageRows
+	})
+
+	manifest := Manifest{
+		Version:     1,
+		GeneratedAt: generatedAt,
+		Tables: []TableManifest{
+			{Name: "guilds", Files: []string{guildFile}, Columns: []string{"id", "name", "icon", "raw_json", "updated_at"}, Rows: 1},
+			{Name: "channels", Files: []string{channelFile}, Columns: []string{"id", "guild_id", "parent_id", "kind", "name", "topic", "position", "is_nsfw", "is_archived", "is_locked", "is_private_thread", "thread_parent_id", "archive_timestamp", "raw_json", "updated_at"}, Rows: 1},
+			{Name: "members", Files: []string{memberFile}, Columns: []string{"guild_id", "user_id", "username", "global_name", "display_name", "nick", "discriminator", "avatar", "bot", "joined_at", "role_ids_json", "raw_json", "updated_at"}, Rows: 1},
+			{Name: "messages", Files: []string{messageFile}, Columns: []string{"id", "guild_id", "channel_id", "author_id", "message_type", "created_at", "edited_at", "deleted_at", "content", "normalized_content", "reply_to_message_id", "pinned", "has_attachments", "raw_json", "updated_at"}, Rows: messageRows},
+		},
+	}
+	body, err := json.MarshalIndent(manifest, "", "  ")
+	require.NoError(t, err)
+	require.NoError(t, os.WriteFile(filepath.Join(repo, ManifestName), append(body, '\n'), 0o600))
+}
+
+func writeJSONLGzip(t *testing.T, repo, rel string, writeRows func(*json.Encoder) int) {
+	t.Helper()
+	path := filepath.Join(repo, filepath.FromSlash(rel))
+	require.NoError(t, os.MkdirAll(filepath.Dir(path), 0o755))
+	file, err := os.Create(path)
+	require.NoError(t, err)
+	gz := gzip.NewWriter(file)
+	enc := json.NewEncoder(gz)
+	rows := writeRows(enc)
+	require.Positive(t, rows)
+	require.NoError(t, gz.Close())
+	require.NoError(t, file.Close())
+}
+
+func syntheticMemoryContent(unique string, size int) string {
+	if size <= len(unique) {
+		return unique
+	}
+	var b strings.Builder
+	b.Grow(size)
+	b.WriteString(unique)
+	for i := 0; b.Len() < size; i++ {
+		_, _ = fmt.Fprintf(&b, " token_%s_%04d", unique, i)
+	}
+	return b.String()[:size]
+}
+
+func envInt(t *testing.T, name string, fallback int) int {
+	t.Helper()
+	raw := strings.TrimSpace(os.Getenv(name))
+	if raw == "" {
+		return fallback
+	}
+	value, err := strconv.Atoi(raw)
+	require.NoError(t, err)
+	require.Positive(t, value)
+	return value
+}

internal/share/share.go

@@ -252,10 +252,11 @@ func Import(ctx context.Context, s *store.Store, opts Options) (Manifest, error)
 func applyImportPragmas(ctx context.Context, db *sql.DB) (func(context.Context) error, error) {
 	// Snapshot imports touch most of the archive. Keep SQLite's crash recovery
 	// enabled; journal_mode=off can leave the live DB malformed if the process
-	// or host dies mid-import.
+	// or host dies mid-import. Keep temporary storage file-backed and bound the
+	// page cache so large imports and FTS rebuilds do not exhaust small hosts.
 	for _, stmt := range []string{
-		`pragma temp_store = memory`,
-		`pragma cache_size = -262144`,
+		`pragma temp_store = file`,
+		`pragma cache_size = -32768`,
 		`pragma journal_mode = wal`,
 		`pragma synchronous = normal`,
 	} {

internal/share/share_test.go

@@ -311,7 +311,7 @@ func TestImportIfChangedInfersLegacyManifestFilesFromGit(t *testing.T) {
 	require.Len(t, results, 1)
 }
 
-func TestApplyImportPragmasKeepCrashRecoveryEnabled(t *testing.T) {
+func TestApplyImportPragmasBoundImportMemory(t *testing.T) {
 	ctx := context.Background()
 	s := seedStore(t, filepath.Join(t.TempDir(), "dst.db"))
 	defer func() { _ = s.Close() }()
@@ -320,6 +320,14 @@ func TestApplyImportPragmasKeepCrashRecoveryEnabled(t *testing.T) {
 	require.NoError(t, err)
 	defer func() { require.NoError(t, restore(ctx)) }()
 
+	var tempStore int
+	require.NoError(t, s.DB().QueryRowContext(ctx, `pragma temp_store`).Scan(&tempStore))
+	require.Equal(t, 1, tempStore, "snapshot imports should use file-backed temporary storage")
+
+	var cacheSize int
+	require.NoError(t, s.DB().QueryRowContext(ctx, `pragma cache_size`).Scan(&cacheSize))
+	require.GreaterOrEqual(t, cacheSize, -65536)
+
 	var journalMode string
 	require.NoError(t, s.DB().QueryRowContext(ctx, `pragma journal_mode`).Scan(&journalMode))
 	require.NotEqual(t, "off", strings.ToLower(journalMode))