[#281] service/audit: Run each SG check in separate routing at PoR

Signed-off-by: Alex Vanin <alexey@nspcc.ru>

pkg/services/audit/auditor/context.go

@@ -12,6 +12,7 @@ import (
 	"github.com/nspcc-dev/neofs-node/pkg/services/object_manager/placement"
 	"github.com/nspcc-dev/neofs-node/pkg/util"
 	"github.com/nspcc-dev/neofs-node/pkg/util/logger"
+	"go.uber.org/atomic"
 	"go.uber.org/zap"
 )
 
@@ -23,13 +24,13 @@ type Context struct {
 
 	report *audit.Report
 
-	// consider adding mutex to access caches
+	sgMembersMtx   sync.RWMutex
-
 	sgMembersCache map[int][]*object.ID
 
+	placementMtx   sync.Mutex
 	placementCache map[string][]netmap.Nodes
 
-	porRequests, porRetries uint32
+	porRequests, porRetries atomic.Uint32
 
 	pairs []gamePair
 
@@ -42,6 +43,7 @@ type Context struct {
 
 	cnrNodesNum int
 
+	headMtx       sync.RWMutex
 	headResponses map[string]shortHeader
 }
 
@@ -130,7 +132,7 @@ func (c *Context) WithTask(t *audit.Task) *Context {
 	return c
 }
 
-// WithPDPWorkerPool sets worker pool for PDP pairs processing..
+// WithPDPWorkerPool sets worker pool for PDP pairs processing.
 func (c *Context) WithPDPWorkerPool(pool util.WorkerPool) *Context {
 	if c != nil {
 		c.pdpWorkerPool = pool
@@ -198,6 +200,9 @@ func (c *Context) writeReport() {
 }
 
 func (c *Context) buildPlacement(id *object.ID) ([]netmap.Nodes, error) {
+	c.placementMtx.Lock()
+	defer c.placementMtx.Unlock()
+
 	strID := id.String()
 
 	if nn, ok := c.placementCache[strID]; ok {
@@ -219,6 +224,9 @@ func (c *Context) buildPlacement(id *object.ID) ([]netmap.Nodes, error) {
 }
 
 func (c *Context) objectSize(id *object.ID) uint64 {
+	c.headMtx.RLock()
+	defer c.headMtx.RUnlock()
+
 	strID := id.String()
 
 	if hdr, ok := c.headResponses[strID]; ok {
@@ -229,6 +237,9 @@ func (c *Context) objectSize(id *object.ID) uint64 {
 }
 
 func (c *Context) objectHomoHash(id *object.ID) []byte {
+	c.headMtx.RLock()
+	defer c.headMtx.RUnlock()
+
 	strID := id.String()
 
 	if hdr, ok := c.headResponses[strID]; ok {
@@ -239,6 +250,9 @@ func (c *Context) objectHomoHash(id *object.ID) []byte {
 }
 
 func (c *Context) updateHeadResponses(hdr *object.Object) {
+	c.headMtx.Lock()
+	defer c.headMtx.Unlock()
+
 	strID := hdr.ID().String()
 
 	if _, ok := c.headResponses[strID]; !ok {
@@ -248,3 +262,10 @@ func (c *Context) updateHeadResponses(hdr *object.Object) {
 		}
 	}
 }
+
+func (c *Context) updateSGInfo(ind int, members []*object.ID) {
+	c.sgMembersMtx.Lock()
+	defer c.sgMembersMtx.Unlock()
+
+	c.sgMembersCache[ind] = members
+}

pkg/services/audit/auditor/pop.go

@@ -162,6 +162,9 @@ func (c *Context) iterateSGMembersRand(f func(*object.ID) bool) {
 }
 
 func (c *Context) iterateSGInfo(f func([]*object.ID) bool) {
+	c.sgMembersMtx.RLock()
+	defer c.sgMembersMtx.RUnlock()
+
 	// we can add randomization like for SG members,
 	// but list of storage groups is already expected
 	// to be shuffled since it is a Search response

pkg/services/audit/auditor/por.go

@@ -2,6 +2,7 @@ package auditor
 
 import (
 	"bytes"
+	"sync"
 
 	"github.com/nspcc-dev/neofs-api-go/pkg/object"
 	"github.com/nspcc-dev/neofs-node/pkg/services/object_manager/placement"
@@ -11,10 +12,25 @@ import (
 )
 
 func (c *Context) executePoR() {
-	for i, sg := range c.task.StorageGroupList() {
+	wg := new(sync.WaitGroup)
-		c.checkStorageGroupPoR(i, sg) // consider parallel it
+	sgs := c.task.StorageGroupList()
+
+	for i := range sgs {
+		wg.Add(1)
+
+		sg := sgs[i]
+
+		if err := c.porWorkerPool.Submit(func() {
+			c.checkStorageGroupPoR(i, sg)
+			wg.Done()
+		}); err != nil {
+			wg.Done()
+		}
 	}
-	c.report.SetPoRCounters(c.porRequests, c.porRetries)
+
+	wg.Wait()
+
+	c.report.SetPoRCounters(c.porRequests.Load(), c.porRetries.Load())
 }
 
 func (c *Context) checkStorageGroupPoR(ind int, sg *object.ID) {
@@ -28,11 +44,13 @@ func (c *Context) checkStorageGroupPoR(ind int, sg *object.ID) {
 	}
 
 	members := storageGroup.Members()
-	c.sgMembersCache[ind] = members
+	c.updateSGInfo(ind, members)
 
 	var (
 		tzHash    []byte
 		totalSize uint64
+
+		accRequests, accRetries uint32
 	)
 
 	for i := range members {
@@ -54,9 +72,9 @@ func (c *Context) checkStorageGroupPoR(ind int, sg *object.ID) {
 		})
 
 		for i := range flat {
-			c.porRequests++
+			accRequests++
 			if i > 0 { // in best case audit get object header on first iteration
-				c.porRetries++
+				accRetries++
 			}
 
 			hdr, err := c.cnrCom.GetHeader(c.task, flat[i], members[i], true)
@@ -93,6 +111,9 @@ func (c *Context) checkStorageGroupPoR(ind int, sg *object.ID) {
 		}
 	}
 
+	c.porRequests.Add(accRequests)
+	c.porRetries.Add(accRetries)
+
 	sizeCheck := storageGroup.ValidationDataSize() == totalSize
 	tzCheck := bytes.Equal(tzHash, storageGroup.ValidationDataHash().Sum())