[#776] writecache: Limit size of used disk space

There is a need to limit disk space used by write-cache. It is almost
impossible to calculate the value exactly. It is proposed to estimate the
size of the cache by the number of objects stored in it.

Track amounts of objects saved in DB and FSTree separately. To do this,
`ObjectCounters` interface is defined. It is generalized to a store of
numbers that can be made persistent (new option `WithObjectCounters`). By
default DB number is calculated as key number in default bucket, and FS
number is set same to DB since it is currently hard to read the actual value
from `FSTree` instance. Each PUT/DELETE operation to DB or FS
increases/decreases corresponding counter. Before each PUT op an overflow
check is performed with the following formula for evaluating the occupied
space: `NumDB * MaxDBSize + NumFS * MaxFSSize`. If next PUT can cause
write-cache overflow, object is written to the main storage.

By default maximum write-cache size is set to 1GB.

Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
This commit is contained in:
Leonard Lyubich 2021-09-08 12:32:20 +03:00 committed by Alex Vanin
parent 0a130177d6
commit a1696a81b6
6 changed files with 203 additions and 10 deletions

View file

@ -47,6 +47,7 @@ func (c *cache) Delete(addr *objectSDK.Address) error {
}
c.dbSize.Sub(uint64(has))
storagelog.Write(c.log, storagelog.AddressField(saddr), storagelog.OpField("db DELETE"))
c.objCounters.DecDB()
return nil
}
@ -57,6 +58,7 @@ func (c *cache) Delete(addr *objectSDK.Address) error {
if err == nil {
storagelog.Write(c.log, storagelog.AddressField(saddr), storagelog.OpField("fstree DELETE"))
c.objCounters.DecFS()
}
return err

View file

@ -26,6 +26,11 @@ type options struct {
smallObjectSize uint64
// workersCount is the number of workers flushing objects in parallel.
workersCount int
// maxCacheSize is the maximum total size of all objects saved in cache (DB + FS).
// 1 GiB by default.
maxCacheSize uint64
// objCounters is an ObjectCounters instance needed for cache size estimation.
objCounters ObjectCounters
}
// WithLogger sets logger.
@ -88,3 +93,17 @@ func WithFlushWorkersCount(c int) Option {
}
}
}
// WithObjectCounters sets ObjectCounters instance needed for cache write-cache size estimation.
func WithObjectCounters(v ObjectCounters) Option {
return func(o *options) {
o.objCounters = v
}
}
// WithMaxCacheSize sets maximum write-cache size in bytes.
func WithMaxCacheSize(sz uint64) Option {
return func(o *options) {
o.maxCacheSize = sz
}
}

View file

@ -54,18 +54,28 @@ func (c *cache) persistLoop() {
func (c *cache) persistToCache(objs []objectInfo) []int {
var (
failMem []int
failMem []int // some index is negative => all objects starting from it will overflow the cache
doneMem []int
)
var sz uint64
err := c.db.Update(func(tx *bbolt.Tx) error {
b := tx.Bucket(defaultBucket)
cacheSz := c.estimateCacheSize()
for i := range objs {
if uint64(len(objs[i].data)) >= c.smallObjectSize {
failMem = append(failMem, i)
continue
}
// check if object will overflow write-cache size limit
updCacheSz := c.incSizeDB(cacheSz)
if updCacheSz > c.maxCacheSize {
// set negative index. We decrement index to cover 0 val (overflow is practically impossible)
failMem = append(failMem, -i-1)
return nil
}
err := b.Put([]byte(objs[i].addr), objs[i].data)
if err != nil {
return err
@ -73,6 +83,10 @@ func (c *cache) persistToCache(objs []objectInfo) []int {
sz += uint64(len(objs[i].data))
doneMem = append(doneMem, i)
storagelog.Write(c.log, storagelog.AddressField(objs[i].addr), storagelog.OpField("db PUT"))
// update cache size
cacheSz = updCacheSz
c.objCounters.IncDB()
}
return nil
})
@ -88,17 +102,55 @@ func (c *cache) persistToCache(objs []objectInfo) []int {
var failDisk []int
for _, i := range failMem {
if uint64(len(objs[i].data)) > c.maxObjectSize {
failDisk = append(failDisk, i)
cacheSz := c.estimateCacheSize()
for _, objInd := range failMem {
var (
updCacheSz uint64
overflowInd = -1
)
if objInd < 0 {
// actually, since the overflow was detected in DB tx, the required space could well have been freed,
// but it is easier to consider the entire method atomic
overflowInd = -objInd - 1 // subtract 1 since we decremented index above
} else {
// check if object will overflow write-cache size limit
if updCacheSz = c.incSizeFS(cacheSz); updCacheSz > c.maxCacheSize {
overflowInd = objInd
}
}
if overflowInd >= 0 {
loop:
for j := range objs[overflowInd:] {
// exclude objects which are already stored in DB
for _, doneMemInd := range doneMem {
if j == doneMemInd {
continue loop
}
}
failDisk = append(failDisk, j)
}
break
}
if uint64(len(objs[objInd].data)) > c.maxObjectSize {
failDisk = append(failDisk, objInd)
continue
}
err := c.fsTree.Put(objs[i].obj.Address(), objs[i].data)
err := c.fsTree.Put(objs[objInd].obj.Address(), objs[objInd].data)
if err != nil {
failDisk = append(failDisk, i)
failDisk = append(failDisk, objInd)
} else {
storagelog.Write(c.log, storagelog.AddressField(objs[i].addr), storagelog.OpField("fstree PUT"))
storagelog.Write(c.log, storagelog.AddressField(objs[objInd].addr), storagelog.OpField("fstree PUT"))
// update cache size
cacheSz = updCacheSz
c.objCounters.IncFS()
}
}

View file

@ -0,0 +1,103 @@
package writecache
import (
"fmt"
"github.com/nspcc-dev/neofs-node/pkg/local_object_storage/blobstor/fstree"
"go.etcd.io/bbolt"
"go.uber.org/atomic"
)
// ObjectCounters is an interface of the storage of cached object amount.
type ObjectCounters interface {
// Increments number of objects saved in DB.
IncDB()
// Decrements number of objects saved in DB.
DecDB()
// Returns number of objects saved in DB.
DB() uint64
// Increments number of objects saved in FSTree.
IncFS()
// Decrements number of objects saved in FSTree.
DecFS()
// Returns number of objects saved in FSTree.
FS() uint64
// Reads number of objects saved in write-cache. It is called on write-cache initialization step.
Read() error
// Flushes the values and closes the storage. It is called on write-cache shutdown.
FlushAndClose()
}
func (c *cache) estimateCacheSize() uint64 {
return c.objCounters.DB()*c.smallObjectSize + c.objCounters.FS()*c.maxObjectSize
}
func (c *cache) incSizeDB(sz uint64) uint64 {
return sz + c.smallObjectSize
}
func (c *cache) incSizeFS(sz uint64) uint64 {
return sz + c.maxObjectSize
}
type counters struct {
cDB, cFS atomic.Uint64
db *bbolt.DB
fs *fstree.FSTree
}
func (x *counters) IncDB() {
x.cDB.Inc()
}
func (x *counters) DecDB() {
x.cDB.Dec()
}
func (x *counters) DB() uint64 {
return x.cDB.Load()
}
func (x *counters) IncFS() {
x.cFS.Inc()
}
func (x *counters) DecFS() {
x.cFS.Dec()
}
func (x *counters) FS() uint64 {
return x.cFS.Load()
}
func (x *counters) Read() error {
var inDB uint64
err := x.db.View(func(tx *bbolt.Tx) error {
b := tx.Bucket(defaultBucket)
if b != nil {
inDB = uint64(b.Stats().KeyN)
}
return nil
})
if err != nil {
return fmt.Errorf("could not read write-cache DB counter: %w", err)
}
x.cDB.Store(inDB)
// FIXME: calculate the actual value in FSTree (new method?).
// For now we can think that db/fs = 50/50.
x.cFS.Store(inDB)
return nil
}
func (x *counters) FlushAndClose() {
// values aren't stored
}

View file

@ -118,6 +118,7 @@ func (c *cache) deleteFromDB(keys [][]byte) error {
return err
}
c.dbSize.Sub(sz)
c.objCounters.DecDB()
return nil
}
@ -139,6 +140,7 @@ func (c *cache) deleteFromDisk(keys [][]byte) error {
continue
} else if err == nil {
storagelog.Write(c.log, storagelog.AddressField(string(keys[i])), storagelog.OpField("fstree DELETE"))
c.objCounters.DecFS()
}
}

View file

@ -60,6 +60,7 @@ const (
maxInMemorySizeBytes = 1024 * 1024 * 1024 // 1 GiB
maxObjectSize = 64 * 1024 * 1024 // 64 MiB
smallObjectSize = 32 * 1024 // 32 KiB
maxCacheSizeBytes = 1 << 30 // 1 GiB
)
var (
@ -81,6 +82,7 @@ func New(opts ...Option) Cache {
maxObjectSize: maxObjectSize,
smallObjectSize: smallObjectSize,
workersCount: flushWorkersCount,
maxCacheSize: maxCacheSizeBytes,
},
}
@ -91,9 +93,21 @@ func New(opts ...Option) Cache {
return c
}
// Open opens and initializes database.
// Open opens and initializes database. Reads object counters from the ObjectCounters instance.
func (c *cache) Open() error {
return c.openStore()
err := c.openStore()
if err != nil {
return err
}
if c.objCounters == nil {
c.objCounters = &counters{
db: c.db,
fs: c.fsTree,
}
}
return c.objCounters.Read()
}
// Init runs necessary services.
@ -103,8 +117,9 @@ func (c *cache) Init() error {
return nil
}
// Close closes db connection and stops services.
// Close closes db connection and stops services. Executes ObjectCounters.FlushAndClose op.
func (c *cache) Close() error {
close(c.closeCh)
c.objCounters.FlushAndClose()
return c.db.Close()
}