lib/db: Slightly improve indirection (ref #6372) (#6373)

I was working on indirecting version vectors, and that resulted in some
refactoring and improving the existing block indirection stuff. We may
or may not end up doing the version vector indirection, but I think
these changes are reasonable anyhow and will simplify the diff
significantly if we do go there. The main points are:

- A bunch of renaming to make the indirection and GC not about "blocks"
  but about "indirection".

- Adding a cutoff so that we don't actually indirect for small block
  lists. This gets us better performance when handling small files as it
  cuts out the indirection for quite small loss in space efficiency.

- Being paranoid and always recalculating the hash on put. This costs
  some CPU, but the consequences if a buggy or malicious implementation
  silently substituted the block list by lying about the hash would be bad.
This commit is contained in:
Jakob Borg 2020-02-27 11:19:21 +01:00
parent a5e12a0a3d
commit ee61da5b6a
5 changed files with 94 additions and 75 deletions

View File

@ -148,7 +148,7 @@ func idxck(ldb backend.Backend) (success bool) {
}
}
if fi.BlocksHash != nil {
if len(fi.Blocks) == 0 && len(fi.BlocksHash) != 0 {
key := string(fi.BlocksHash)
if _, ok := blocklists[key]; !ok {
fmt.Printf("Missing block list for file %q, block list hash %x\n", fi.Name, fi.BlocksHash)

View File

@ -119,8 +119,8 @@ are mostly useful for developers. Use with care.
"h", "m" and "s" abbreviations for hours minutes and seconds.
Valid values are like "720h", "30s", etc.
STGCBLOCKSEVERY Set to a time interval to override the default database
block GC interval of 13 hours. Same format as the
STGCINDIRECTEVERY Set to a time interval to override the default database
indirection GC interval of 13 hours. Same format as the
STRECHECKDBEVERY variable.
GOMAXPROCS Set the maximum number of CPU cores to use. Defaults to all

View File

@ -19,22 +19,30 @@ import (
)
const (
// We set the bloom filter capacity to handle 100k individual block lists
// with a false positive probability of 1% for the first pass. Once we know
// how many block lists we have we will use that number instead, if it's
// more than 100k. For fewer than 100k block lists we will just get better
// false positive rate instead.
blockGCBloomCapacity = 100000
blockGCBloomFalsePositiveRate = 0.01 // 1%
blockGCDefaultInterval = 13 * time.Hour
blockGCTimeKey = "lastBlockGCTime"
// We set the bloom filter capacity to handle 100k individual items with
// a false positive probability of 1% for the first pass. Once we know
// how many items we have we will use that number instead, if it's more
// than 100k. For fewer than 100k items we will just get better false
// positive rate instead.
indirectGCBloomCapacity = 100000
indirectGCBloomFalsePositiveRate = 0.01 // 1%
indirectGCDefaultInterval = 13 * time.Hour
indirectGCTimeKey = "lastIndirectGCTime"
// Use indirection for the block list when it exceeds this many entries
blocksIndirectionCutoff = 3
)
var blockGCInterval = blockGCDefaultInterval
var indirectGCInterval = indirectGCDefaultInterval
func init() {
// deprecated
if dur, err := time.ParseDuration(os.Getenv("STGCBLOCKSEVERY")); err == nil {
blockGCInterval = dur
indirectGCInterval = dur
}
// current
if dur, err := time.ParseDuration(os.Getenv("STGCINDIRECTEVERY")); err == nil {
indirectGCInterval = dur
}
}
@ -485,18 +493,18 @@ func (db *Lowlevel) dropPrefix(prefix []byte) error {
}
func (db *Lowlevel) gcRunner() {
t := time.NewTimer(db.timeUntil(blockGCTimeKey, blockGCInterval))
t := time.NewTimer(db.timeUntil(indirectGCTimeKey, indirectGCInterval))
defer t.Stop()
for {
select {
case <-db.gcStop:
return
case <-t.C:
if err := db.gcBlocks(); err != nil {
l.Warnln("Database block GC failed:", err)
if err := db.gcIndirect(); err != nil {
l.Warnln("Database indirection GC failed:", err)
}
db.recordTime(blockGCTimeKey)
t.Reset(db.timeUntil(blockGCTimeKey, blockGCInterval))
db.recordTime(indirectGCTimeKey)
t.Reset(db.timeUntil(indirectGCTimeKey, indirectGCInterval))
}
}
}
@ -521,15 +529,16 @@ func (db *Lowlevel) timeUntil(key string, every time.Duration) time.Duration {
return sleepTime
}
func (db *Lowlevel) gcBlocks() error {
// The block GC uses a bloom filter to track used block lists. This means
// iterating over all items, adding their block lists to the filter, then
// iterating over the block lists and removing those that don't match the
// filter. The filter will give false positives so we will keep around one
// percent of block lists that we don't really need (at most).
func (db *Lowlevel) gcIndirect() error {
// The indirection GC uses bloom filters to track used block lists and
// versions. This means iterating over all items, adding their hashes to
// the filter, then iterating over the indirected items and removing
// those that don't match the filter. The filter will give false
// positives so we will keep around one percent of things that we don't
// really need (at most).
//
// Block GC needs to run when there are no modifications to the FileInfos or
// block lists.
// Indirection GC needs to run when there are no modifications to the
// FileInfos or indirected items.
db.gcMut.Lock()
defer db.gcMut.Unlock()
@ -540,18 +549,19 @@ func (db *Lowlevel) gcBlocks() error {
}
defer t.Release()
// Set up the bloom filter with the initial capacity and false positive
// rate, or higher capacity if we've done this before and seen lots of block
// lists.
// Set up the bloom filters with the initial capacity and false positive
// rate, or higher capacity if we've done this before and seen lots of
// items. For simplicity's sake we track just one count, which is the
// highest of the various indirected items.
capacity := blockGCBloomCapacity
capacity := indirectGCBloomCapacity
if db.gcKeyCount > capacity {
capacity = db.gcKeyCount
}
filter := bloom.NewWithEstimates(uint(capacity), blockGCBloomFalsePositiveRate)
blockFilter := bloom.NewWithEstimates(uint(capacity), indirectGCBloomFalsePositiveRate)
// Iterate the FileInfos, unmarshal the blocks hashes and add them to
// the filter.
// Iterate the FileInfos, unmarshal the block and version hashes and
// add them to the filter.
it, err := db.NewPrefixIterator([]byte{KeyTypeDevice})
if err != nil {
@ -563,7 +573,7 @@ func (db *Lowlevel) gcBlocks() error {
return err
}
if len(bl.BlocksHash) > 0 {
filter.Add(bl.BlocksHash)
blockFilter.Add(bl.BlocksHash)
}
}
it.Release()
@ -578,11 +588,11 @@ func (db *Lowlevel) gcBlocks() error {
if err != nil {
return err
}
matched := 0
matchedBlocks := 0
for it.Next() {
key := blockListKey(it.Key())
if filter.Test(key.BlocksHash()) {
matched++
if blockFilter.Test(key.BlocksHash()) {
matchedBlocks++
continue
}
if err := t.Delete(key); err != nil {
@ -595,7 +605,7 @@ func (db *Lowlevel) gcBlocks() error {
}
// Remember the number of unique keys we kept until the next pass.
db.gcKeyCount = matched
db.gcKeyCount = matchedBlocks
if err := t.Commit(); err != nil {
return err

View File

@ -446,10 +446,11 @@ func (db *schemaUpdater) updateSchemato9(prev int) error {
}
metas := make(map[string]*metadataTracker)
for it.Next() {
var fi protocol.FileInfo
if err := fi.Unmarshal(it.Value()); err != nil {
intf, err := t.unmarshalTrunc(it.Value(), false)
if err != nil {
return err
}
fi := intf.(protocol.FileInfo)
device, ok := t.keyer.DeviceFromDeviceFileKey(it.Key())
if !ok {
return errDeviceIdxMissing
@ -510,7 +511,7 @@ func (db *schemaUpdater) updateSchemato9(prev int) error {
}
}
db.recordTime(blockGCTimeKey)
db.recordTime(indirectGCTimeKey)
return t.Commit()
}

View File

@ -78,22 +78,24 @@ func (t readOnlyTransaction) unmarshalTrunc(bs []byte, trunc bool) (FileIntf, er
return tf, nil
}
var tf protocol.FileInfo
if err := tf.Unmarshal(bs); err != nil {
var fi protocol.FileInfo
if err := fi.Unmarshal(bs); err != nil {
return nil, err
}
if err := t.fillBlockList(&tf); err != nil {
if err := t.fillFileInfo(&fi); err != nil {
return nil, err
}
return tf, nil
return fi, nil
}
func (t readOnlyTransaction) fillBlockList(fi *protocol.FileInfo) error {
if len(fi.BlocksHash) == 0 {
return nil
}
blocksKey := t.keyer.GenerateBlockListKey(nil, fi.BlocksHash)
bs, err := t.Get(blocksKey)
// fillFileInfo follows the (possible) indirection of blocks and fills it out.
func (t readOnlyTransaction) fillFileInfo(fi *protocol.FileInfo) error {
var key []byte
if len(fi.Blocks) == 0 && len(fi.BlocksHash) != 0 {
// The blocks list is indirected and we need to load it.
key = t.keyer.GenerateBlockListKey(key, fi.BlocksHash)
bs, err := t.Get(key)
if err != nil {
return err
}
@ -102,6 +104,8 @@ func (t readOnlyTransaction) fillBlockList(fi *protocol.FileInfo) error {
return err
}
fi.Blocks = bl.Blocks
}
return nil
}
@ -453,26 +457,33 @@ func (t readWriteTransaction) close() {
t.WriteTransaction.Release()
}
func (t readWriteTransaction) putFile(key []byte, fi protocol.FileInfo) error {
if fi.Blocks != nil {
if fi.BlocksHash == nil {
func (t readWriteTransaction) putFile(fkey []byte, fi protocol.FileInfo) error {
var bkey []byte
// Always set the blocks hash when there are blocks.
if len(fi.Blocks) > 0 {
fi.BlocksHash = protocol.BlocksHash(fi.Blocks)
} else {
fi.BlocksHash = nil
}
blocksKey := t.keyer.GenerateBlockListKey(nil, fi.BlocksHash)
if _, err := t.Get(blocksKey); backend.IsNotFound(err) {
// Indirect the blocks if the block list is large enough.
if len(fi.Blocks) > blocksIndirectionCutoff {
bkey = t.keyer.GenerateBlockListKey(bkey, fi.BlocksHash)
if _, err := t.Get(bkey); backend.IsNotFound(err) {
// Marshal the block list and save it
blocksBs := mustMarshal(&BlockList{Blocks: fi.Blocks})
if err := t.Put(blocksKey, blocksBs); err != nil {
if err := t.Put(bkey, blocksBs); err != nil {
return err
}
} else if err != nil {
return err
}
fi.Blocks = nil
}
fi.Blocks = nil
fiBs := mustMarshal(&fi)
return t.Put(key, fiBs)
return t.Put(fkey, fiBs)
}
// updateGlobal adds this device+version to the version list for the given
@ -723,15 +734,12 @@ func (t *readWriteTransaction) withAllFolderTruncated(folder []byte, fn func(dev
}
continue
}
var f FileInfoTruncated
// The iterator function may keep a reference to the unmarshalled
// struct, which in turn references the buffer it was unmarshalled
// from. dbi.Value() just returns an internal slice that it reuses, so
// we need to copy it.
err := f.Unmarshal(append([]byte{}, dbi.Value()...))
intf, err := t.unmarshalTrunc(dbi.Value(), true)
if err != nil {
return err
}
f := intf.(FileInfoTruncated)
switch f.Name {
case "", ".", "..", "/": // A few obviously invalid filenames