Merge pull request #6833 from filecoin-project/feat/splitstore-badger…

…-markset

Splitstore: support on-disk marksets using badger

blockstore/splitstore/README.md

@@ -27,9 +27,38 @@ If you intend to use the discard coldstore, your also need to add the following:
     ColdStoreType = "discard"
 ```
 In general you _should not_ have to use the discard store, unless you
-are running a network booster or have very constrained hardware with
-not enough disk space to maintain a coldstore, even with garbage
-collection.
+are running a network assistive node (like a bootstrapper or booster)
+or have very constrained hardware with not enough disk space to
+maintain a coldstore, even with garbage collection. It is also appropriate
+for small nodes that are simply watching the chain.
+
+*Warning:* Using the discard store for a general purpose node is discouraged, unless
+you really know what you are doing. Use it at your own risk.
+
+## Configuration Options
+
+These are options in the `[Chainstore.Splitstore]` section of the configuration:
+
+- `HotStoreType` -- specifies the type of hotstore to use.
+  The only currently supported option is `"badger"`.
+- `ColdStoreType` -- specifies the type of coldstore to use.
+  The default value is `"universal"`, which will use the initial monolith blockstore
+  as the coldstore.
+  The other possible value is `"discard"`, as outlined above, which is specialized for
+  running without a coldstore. Note that the discard store wraps the initial monolith
+  blockstore and discards writes; this is necessary to support syncing from a snapshot.
+- `MarkSetType` -- specifies the type of markset to use during compaction.
+  The markset is the data structure used by compaction/gc to track live objects.
+  The default value is `"map"`, which will use an in-memory map; if you are limited
+  in memory (or indeed see compaction run out of memory), you can also specify
+  `"badger"` which will use an disk backed markset, using badger. This will use
+  much less memory, but will also make compaction slower.
+- `HotStoreMessageRetention` -- specifies how many finalities, beyond the 4
+  finalities maintained by default, to maintain messages and message receipts in the
+  hotstore. This is useful for assistive nodes that want to support syncing for other
+  nodes beyond 4 finalities, while running with the discard coldstore option.
+  It is also useful for miners who accept deals and need to lookback messages beyond
+  the 4 finalities, which would otherwise hit the coldstore.
 
 
 ## Operation
@@ -67,6 +96,6 @@ Compaction works transactionally with the following algorithm:
   - We delete in small batches taking a lock; each batch is checked again for marks, from the concurrent transactional mark, so as to never delete anything live
 - We then end the transaction and compact/gc the hotstore.
 
-## Coldstore Garbage Collection
+## Garbage Collection
 
 TBD -- see [#6577](https://github.com/filecoin-project/lotus/issues/6577)

blockstore/splitstore/markset.go

@@ -32,6 +32,8 @@ func OpenMarkSetEnv(path string, mtype string) (MarkSetEnv, error) {
 		return NewBloomMarkSetEnv()
 	case "map":
 		return NewMapMarkSetEnv()
+	case "badger":
+		return NewBadgerMarkSetEnv(path)
 	default:
 		return nil, xerrors.Errorf("unknown mark set type %s", mtype)
 	}

blockstore/splitstore/markset_badger.go

@@ -0,0 +1,230 @@
+package splitstore
+
+import (
+	"os"
+	"path/filepath"
+	"sync"
+
+	"golang.org/x/xerrors"
+
+	"github.com/dgraph-io/badger/v2"
+	"github.com/dgraph-io/badger/v2/options"
+	"go.uber.org/zap"
+
+	cid "github.com/ipfs/go-cid"
+)
+
+type BadgerMarkSetEnv struct {
+	path string
+}
+
+var _ MarkSetEnv = (*BadgerMarkSetEnv)(nil)
+
+type BadgerMarkSet struct {
+	mx      sync.RWMutex
+	cond    sync.Cond
+	pend    map[string]struct{}
+	writing map[int]map[string]struct{}
+	writers int
+	seqno   int
+
+	db   *badger.DB
+	path string
+}
+
+var _ MarkSet = (*BadgerMarkSet)(nil)
+
+var badgerMarkSetBatchSize = 16384
+
+func NewBadgerMarkSetEnv(path string) (MarkSetEnv, error) {
+	msPath := filepath.Join(path, "markset.badger")
+	err := os.MkdirAll(msPath, 0755) //nolint:gosec
+	if err != nil {
+		return nil, xerrors.Errorf("error creating markset directory: %w", err)
+	}
+
+	return &BadgerMarkSetEnv{path: msPath}, nil
+}
+
+func (e *BadgerMarkSetEnv) Create(name string, sizeHint int64) (MarkSet, error) {
+	path := filepath.Join(e.path, name)
+
+	// clean up first
+	err := os.RemoveAll(path)
+	if err != nil {
+		return nil, xerrors.Errorf("error clearing markset directory: %w", err)
+	}
+
+	err = os.MkdirAll(path, 0755) //nolint:gosec
+	if err != nil {
+		return nil, xerrors.Errorf("error creating markset directory: %w", err)
+	}
+
+	opts := badger.DefaultOptions(path)
+	opts.SyncWrites = false
+	opts.CompactL0OnClose = false
+	opts.Compression = options.None
+	// Note: We use FileIO for loading modes to avoid memory thrashing and interference
+	//       between the system blockstore and the markset.
+	//       It was observed that using the default memory mapped option resulted in
+	//       significant interference and unacceptably high block validation times once the markset
+	//       exceeded 1GB in size.
+	opts.TableLoadingMode = options.FileIO
+	opts.ValueLogLoadingMode = options.FileIO
+	opts.Logger = &badgerLogger{
+		SugaredLogger: log.Desugar().WithOptions(zap.AddCallerSkip(1)).Sugar(),
+		skip2:         log.Desugar().WithOptions(zap.AddCallerSkip(2)).Sugar(),
+	}
+
+	db, err := badger.Open(opts)
+	if err != nil {
+		return nil, xerrors.Errorf("error creating badger markset: %w", err)
+	}
+
+	ms := &BadgerMarkSet{
+		pend:    make(map[string]struct{}),
+		writing: make(map[int]map[string]struct{}),
+		db:      db,
+		path:    path,
+	}
+	ms.cond.L = &ms.mx
+
+	return ms, nil
+}
+
+func (e *BadgerMarkSetEnv) Close() error {
+	return os.RemoveAll(e.path)
+}
+
+func (s *BadgerMarkSet) Mark(c cid.Cid) error {
+	s.mx.Lock()
+
+	if s.pend == nil {
+		s.mx.Unlock()
+		return errMarkSetClosed
+	}
+
+	s.pend[string(c.Hash())] = struct{}{}
+
+	if len(s.pend) < badgerMarkSetBatchSize {
+		s.mx.Unlock()
+		return nil
+	}
+
+	pend := s.pend
+	seqno := s.seqno
+	s.seqno++
+	s.writing[seqno] = pend
+	s.pend = make(map[string]struct{})
+	s.writers++
+	s.mx.Unlock()
+
+	defer func() {
+		s.mx.Lock()
+		defer s.mx.Unlock()
+
+		delete(s.writing, seqno)
+		s.writers--
+		if s.writers == 0 {
+			s.cond.Broadcast()
+		}
+	}()
+
+	empty := []byte{} // not nil
+
+	batch := s.db.NewWriteBatch()
+	defer batch.Cancel()
+
+	for k := range pend {
+		if err := batch.Set([]byte(k), empty); err != nil {
+			return err
+		}
+	}
+
+	err := batch.Flush()
+	if err != nil {
+		return xerrors.Errorf("error flushing batch to badger markset: %w", err)
+	}
+
+	return nil
+}
+
+func (s *BadgerMarkSet) Has(c cid.Cid) (bool, error) {
+	s.mx.RLock()
+	defer s.mx.RUnlock()
+
+	if s.pend == nil {
+		return false, errMarkSetClosed
+	}
+
+	key := c.Hash()
+	pendKey := string(key)
+	_, ok := s.pend[pendKey]
+	if ok {
+		return true, nil
+	}
+
+	for _, wr := range s.writing {
+		_, ok := wr[pendKey]
+		if ok {
+			return true, nil
+		}
+	}
+
+	err := s.db.View(func(txn *badger.Txn) error {
+		_, err := txn.Get(key)
+		return err
+	})
+
+	switch err {
+	case nil:
+		return true, nil
+
+	case badger.ErrKeyNotFound:
+		return false, nil
+
+	default:
+		return false, xerrors.Errorf("error checking badger markset: %w", err)
+	}
+}
+
+func (s *BadgerMarkSet) Close() error {
+	s.mx.Lock()
+	defer s.mx.Unlock()
+
+	if s.pend == nil {
+		return nil
+	}
+
+	for s.writers > 0 {
+		s.cond.Wait()
+	}
+
+	s.pend = nil
+	db := s.db
+	s.db = nil
+
+	err := db.Close()
+	if err != nil {
+		return xerrors.Errorf("error closing badger markset: %w", err)
+	}
+
+	err = os.RemoveAll(s.path)
+	if err != nil {
+		return xerrors.Errorf("error deleting badger markset: %w", err)
+	}
+
+	return nil
+}
+
+func (s *BadgerMarkSet) SetConcurrent() {}
+
+// badger logging through go-log
+type badgerLogger struct {
+	*zap.SugaredLogger
+	skip2 *zap.SugaredLogger
+}
+
+func (b *badgerLogger) Warningf(format string, args ...interface{}) {}
+func (b *badgerLogger) Infof(format string, args ...interface{})    {}
+func (b *badgerLogger) Debugf(format string, args ...interface{})   {}

blockstore/splitstore/markset_test.go

@@ -16,6 +16,15 @@ func TestBloomMarkSet(t *testing.T) {
 	testMarkSet(t, "bloom")
 }
 
+func TestBadgerMarkSet(t *testing.T) {
+	bs := badgerMarkSetBatchSize
+	badgerMarkSetBatchSize = 1
+	t.Cleanup(func() {
+		badgerMarkSetBatchSize = bs
+	})
+	testMarkSet(t, "badger")
+}
+
 func testMarkSet(t *testing.T, lsType string) {
 	t.Helper()
 

blockstore/splitstore/splitstore.go

@@ -62,8 +62,11 @@ func init() {
 type Config struct {
 	// MarkSetType is the type of mark set to use.
 	//
-	// Only current sane value is "map", but we may add an option for a disk-backed
-	// markset for memory-constrained situations.
+	// The default value is "map", which uses an in-memory map-backed markset.
+	// If you are constrained in memory (i.e. compaction runs out of memory), you
+	// can use "badger", which will use a disk-backed markset using badger.
+	// Note that compaction will take quite a bit longer when using the "badger" option,
+	// but that shouldn't really matter (as long as it is under 7.5hrs).
 	MarkSetType string
 
 	// DiscardColdBlocks indicates whether to skip moving cold blocks to the coldstore.

blockstore/splitstore/splitstore_compact.go

@@ -184,16 +184,6 @@ func (s *SplitStore) trackTxnRef(c cid.Cid) {
 		return
 	}
 
-	if s.txnProtect != nil {
-		mark, err := s.txnProtect.Has(c)
-		if err != nil {
-			log.Warnf("error checking markset: %s", err)
-			// track it anyways
-		} else if mark {
-			return
-		}
-	}
-
 	s.txnRefsMx.Lock()
 	s.txnRefs[c] = struct{}{}
 	s.txnRefsMx.Unlock()
@@ -209,27 +199,11 @@ func (s *SplitStore) trackTxnRefMany(cids []cid.Cid) {
 	s.txnRefsMx.Lock()
 	defer s.txnRefsMx.Unlock()
 
-	quiet := false
 	for _, c := range cids {
 		if isUnitaryObject(c) {
 			continue
 		}
 
-		if s.txnProtect != nil {
-			mark, err := s.txnProtect.Has(c)
-			if err != nil {
-				if !quiet {
-					quiet = true
-					log.Warnf("error checking markset: %s", err)
-				}
-				// track it anyways
-			}
-
-			if mark {
-				continue
-			}
-		}
-
 		s.txnRefs[c] = struct{}{}
 	}
 

blockstore/splitstore/splitstore_test.go

@@ -210,6 +210,15 @@ func TestSplitStoreCompaction(t *testing.T) {
 	testSplitStore(t, &Config{MarkSetType: "map"})
 }
 
+func TestSplitStoreCompactionWithBadger(t *testing.T) {
+	bs := badgerMarkSetBatchSize
+	badgerMarkSetBatchSize = 1
+	t.Cleanup(func() {
+		badgerMarkSetBatchSize = bs
+	})
+	testSplitStore(t, &Config{MarkSetType: "badger"})
+}
+
 type mockChain struct {
 	t testing.TB