forked from TrueCloudLab/frostfs-node
[#248] services: Remove audit service
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
This commit is contained in:
parent
8879c6ea4a
commit
7f49f07255
12 changed files with 0 additions and 1358 deletions
|
@ -1,298 +0,0 @@
|
|||
package auditor
|
||||
|
||||
import (
|
||||
"context"
|
||||
"sync"
|
||||
"time"
|
||||
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/internal/logs"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/services/audit"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/services/object_manager/placement"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/util"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/util/logger"
|
||||
cid "git.frostfs.info/TrueCloudLab/frostfs-sdk-go/container/id"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-sdk-go/netmap"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-sdk-go/object"
|
||||
oid "git.frostfs.info/TrueCloudLab/frostfs-sdk-go/object/id"
|
||||
"go.uber.org/atomic"
|
||||
"go.uber.org/zap"
|
||||
)
|
||||
|
||||
// Context represents container data audit execution context.
|
||||
type Context struct {
|
||||
ContextPrm
|
||||
|
||||
task *audit.Task
|
||||
|
||||
report *audit.Report
|
||||
|
||||
sgMembersMtx sync.RWMutex
|
||||
sgMembersCache map[oid.ID][]oid.ID
|
||||
|
||||
placementMtx sync.Mutex
|
||||
placementCache map[string][][]netmap.NodeInfo
|
||||
|
||||
porRequests, porRetries atomic.Uint32
|
||||
|
||||
pairs []gamePair
|
||||
|
||||
pairedMtx sync.Mutex
|
||||
pairedNodes map[uint64]*pairMemberInfo
|
||||
|
||||
counters struct {
|
||||
hit, miss, fail uint32
|
||||
}
|
||||
|
||||
cnrNodesNum int
|
||||
|
||||
headMtx sync.RWMutex
|
||||
headResponses map[string]shortHeader
|
||||
}
|
||||
|
||||
type pairMemberInfo struct {
|
||||
failedPDP, passedPDP bool // at least one
|
||||
|
||||
node netmap.NodeInfo
|
||||
}
|
||||
|
||||
type gamePair struct {
|
||||
n1, n2 netmap.NodeInfo
|
||||
|
||||
id oid.ID
|
||||
|
||||
rn1, rn2 []*object.Range
|
||||
|
||||
hh1, hh2 [][]byte
|
||||
}
|
||||
|
||||
type shortHeader struct {
|
||||
tzhash []byte
|
||||
|
||||
objectSize uint64
|
||||
}
|
||||
|
||||
// ContextPrm groups components required to conduct data audit checks.
|
||||
type ContextPrm struct {
|
||||
maxPDPSleep uint64
|
||||
|
||||
log *logger.Logger
|
||||
|
||||
cnrCom ContainerCommunicator
|
||||
|
||||
pdpWorkerPool, porWorkerPool util.WorkerPool
|
||||
}
|
||||
|
||||
type commonCommunicatorPrm struct {
|
||||
Node netmap.NodeInfo
|
||||
|
||||
OID oid.ID
|
||||
CID cid.ID
|
||||
}
|
||||
|
||||
// GetHeaderPrm groups parameter of GetHeader operation.
|
||||
type GetHeaderPrm struct {
|
||||
commonCommunicatorPrm
|
||||
|
||||
NodeIsRelay bool
|
||||
}
|
||||
|
||||
// GetRangeHashPrm groups parameter of GetRangeHash operation.
|
||||
type GetRangeHashPrm struct {
|
||||
commonCommunicatorPrm
|
||||
|
||||
Range *object.Range
|
||||
}
|
||||
|
||||
// ContainerCommunicator is an interface of
|
||||
// component of communication with container nodes.
|
||||
type ContainerCommunicator interface {
|
||||
// GetHeader must return object header from the container node.
|
||||
GetHeader(context.Context, GetHeaderPrm) (*object.Object, error)
|
||||
|
||||
// GetRangeHash must return homomorphic Tillich-Zemor hash of payload range of the
|
||||
// object stored in container node.
|
||||
GetRangeHash(context.Context, GetRangeHashPrm) ([]byte, error)
|
||||
}
|
||||
|
||||
// NewContext creates, initializes and returns Context.
|
||||
func NewContext(prm ContextPrm) *Context {
|
||||
return &Context{
|
||||
ContextPrm: prm,
|
||||
}
|
||||
}
|
||||
|
||||
// SetLogger sets logging component.
|
||||
func (p *ContextPrm) SetLogger(l *logger.Logger) {
|
||||
if p != nil {
|
||||
p.log = l
|
||||
}
|
||||
}
|
||||
|
||||
// SetContainerCommunicator sets component of communication with container nodes.
|
||||
func (p *ContextPrm) SetContainerCommunicator(cnrCom ContainerCommunicator) {
|
||||
if p != nil {
|
||||
p.cnrCom = cnrCom
|
||||
}
|
||||
}
|
||||
|
||||
// SetMaxPDPSleep sets maximum sleep interval between range hash requests.
|
||||
// as part of PDP check.
|
||||
func (p *ContextPrm) SetMaxPDPSleep(dur time.Duration) {
|
||||
if p != nil {
|
||||
p.maxPDPSleep = uint64(dur)
|
||||
}
|
||||
}
|
||||
|
||||
// WithTask sets container audit parameters.
|
||||
func (c *Context) WithTask(t *audit.Task) *Context {
|
||||
if c != nil {
|
||||
c.task = t
|
||||
}
|
||||
|
||||
return c
|
||||
}
|
||||
|
||||
// WithPDPWorkerPool sets worker pool for PDP pairs processing.
|
||||
func (c *Context) WithPDPWorkerPool(pool util.WorkerPool) *Context {
|
||||
if c != nil {
|
||||
c.pdpWorkerPool = pool
|
||||
}
|
||||
|
||||
return c
|
||||
}
|
||||
|
||||
// WithPoRWorkerPool sets worker pool for PoR SG processing.
|
||||
func (c *Context) WithPoRWorkerPool(pool util.WorkerPool) *Context {
|
||||
if c != nil {
|
||||
c.porWorkerPool = pool
|
||||
}
|
||||
|
||||
return c
|
||||
}
|
||||
|
||||
func (c *Context) containerID() cid.ID {
|
||||
return c.task.ContainerID()
|
||||
}
|
||||
|
||||
func (c *Context) init() {
|
||||
c.report = audit.NewReport(c.containerID())
|
||||
|
||||
c.sgMembersCache = make(map[oid.ID][]oid.ID)
|
||||
|
||||
c.placementCache = make(map[string][][]netmap.NodeInfo)
|
||||
|
||||
cnrVectors := c.task.ContainerNodes()
|
||||
for i := range cnrVectors {
|
||||
c.cnrNodesNum += len(cnrVectors[i])
|
||||
}
|
||||
|
||||
c.pairedNodes = make(map[uint64]*pairMemberInfo)
|
||||
|
||||
c.headResponses = make(map[string]shortHeader)
|
||||
|
||||
c.log = &logger.Logger{Logger: c.log.With(
|
||||
zap.Stringer("container ID", c.task.ContainerID()),
|
||||
)}
|
||||
}
|
||||
|
||||
func (c *Context) expired(ctx context.Context) bool {
|
||||
select {
|
||||
case <-ctx.Done():
|
||||
c.log.Debug(logs.AuditorAuditContextIsDone,
|
||||
zap.String("error", ctx.Err().Error()),
|
||||
)
|
||||
|
||||
return true
|
||||
default:
|
||||
return false
|
||||
}
|
||||
}
|
||||
|
||||
func (c *Context) complete() {
|
||||
c.report.Complete()
|
||||
}
|
||||
|
||||
func (c *Context) writeReport() {
|
||||
c.log.Debug(logs.AuditorWritingAuditReport)
|
||||
|
||||
if err := c.task.Reporter().WriteReport(c.report); err != nil {
|
||||
c.log.Error(logs.AuditorCouldNotWriteAuditReport)
|
||||
}
|
||||
}
|
||||
|
||||
func (c *Context) buildPlacement(id oid.ID) ([][]netmap.NodeInfo, error) {
|
||||
c.placementMtx.Lock()
|
||||
defer c.placementMtx.Unlock()
|
||||
|
||||
strID := id.EncodeToString()
|
||||
|
||||
if nn, ok := c.placementCache[strID]; ok {
|
||||
return nn, nil
|
||||
}
|
||||
|
||||
nn, err := placement.BuildObjectPlacement(
|
||||
c.task.NetworkMap(),
|
||||
c.task.ContainerNodes(),
|
||||
&id,
|
||||
)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
c.placementCache[strID] = nn
|
||||
|
||||
return nn, nil
|
||||
}
|
||||
|
||||
func (c *Context) objectSize(id oid.ID) uint64 {
|
||||
c.headMtx.RLock()
|
||||
defer c.headMtx.RUnlock()
|
||||
|
||||
strID := id.EncodeToString()
|
||||
|
||||
if hdr, ok := c.headResponses[strID]; ok {
|
||||
return hdr.objectSize
|
||||
}
|
||||
|
||||
return 0
|
||||
}
|
||||
|
||||
func (c *Context) objectHomoHash(id oid.ID) []byte {
|
||||
c.headMtx.RLock()
|
||||
defer c.headMtx.RUnlock()
|
||||
|
||||
strID := id.EncodeToString()
|
||||
|
||||
if hdr, ok := c.headResponses[strID]; ok {
|
||||
return hdr.tzhash
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
func (c *Context) updateHeadResponses(hdr *object.Object) {
|
||||
id, ok := hdr.ID()
|
||||
if !ok {
|
||||
return
|
||||
}
|
||||
|
||||
strID := id.EncodeToString()
|
||||
cs, _ := hdr.PayloadHomomorphicHash()
|
||||
|
||||
c.headMtx.Lock()
|
||||
defer c.headMtx.Unlock()
|
||||
|
||||
if _, ok := c.headResponses[strID]; !ok {
|
||||
c.headResponses[strID] = shortHeader{
|
||||
tzhash: cs.Value(),
|
||||
objectSize: hdr.PayloadSize(),
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (c *Context) updateSGInfo(id oid.ID, members []oid.ID) {
|
||||
c.sgMembersMtx.Lock()
|
||||
defer c.sgMembersMtx.Unlock()
|
||||
|
||||
c.sgMembersCache[id] = members
|
||||
}
|
|
@ -1,37 +0,0 @@
|
|||
package auditor
|
||||
|
||||
import (
|
||||
"context"
|
||||
"fmt"
|
||||
)
|
||||
|
||||
// Execute audits container data.
|
||||
func (c *Context) Execute(ctx context.Context, onCompleted func()) {
|
||||
defer onCompleted()
|
||||
c.init()
|
||||
|
||||
checks := []struct {
|
||||
name string
|
||||
exec func(context.Context)
|
||||
}{
|
||||
{name: "PoR", exec: c.executePoR},
|
||||
{name: "PoP", exec: c.executePoP},
|
||||
{name: "PDP", exec: c.executePDP},
|
||||
}
|
||||
|
||||
for i := range checks {
|
||||
c.log.Debug(fmt.Sprintf("executing %s check...", checks[i].name))
|
||||
|
||||
if c.expired(ctx) {
|
||||
break
|
||||
}
|
||||
|
||||
checks[i].exec(ctx)
|
||||
|
||||
if i == len(checks)-1 {
|
||||
c.complete()
|
||||
}
|
||||
}
|
||||
|
||||
c.writeReport()
|
||||
}
|
|
@ -1,240 +0,0 @@
|
|||
package auditor
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"context"
|
||||
"sync"
|
||||
"time"
|
||||
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/internal/logs"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/util/rand"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-sdk-go/netmap"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-sdk-go/object"
|
||||
oid "git.frostfs.info/TrueCloudLab/frostfs-sdk-go/object/id"
|
||||
"git.frostfs.info/TrueCloudLab/tzhash/tz"
|
||||
"go.uber.org/zap"
|
||||
)
|
||||
|
||||
func (c *Context) executePDP(ctx context.Context) {
|
||||
c.processPairs(ctx)
|
||||
c.writePairsResult()
|
||||
}
|
||||
|
||||
func (c *Context) processPairs(ctx context.Context) {
|
||||
wg := new(sync.WaitGroup)
|
||||
|
||||
for i := range c.pairs {
|
||||
p := &c.pairs[i]
|
||||
wg.Add(1)
|
||||
|
||||
if err := c.pdpWorkerPool.Submit(func() {
|
||||
c.processPair(ctx, p)
|
||||
wg.Done()
|
||||
}); err != nil {
|
||||
wg.Done()
|
||||
}
|
||||
}
|
||||
|
||||
wg.Wait()
|
||||
c.pdpWorkerPool.Release()
|
||||
}
|
||||
|
||||
func (c *Context) processPair(ctx context.Context, p *gamePair) {
|
||||
c.distributeRanges(p)
|
||||
c.collectHashes(ctx, p)
|
||||
c.analyzeHashes(p)
|
||||
}
|
||||
|
||||
func (c *Context) distributeRanges(p *gamePair) {
|
||||
p.rn1 = make([]*object.Range, hashRangeNumber-1)
|
||||
p.rn2 = make([]*object.Range, hashRangeNumber-1)
|
||||
|
||||
for i := 0; i < hashRangeNumber-1; i++ {
|
||||
p.rn1[i] = object.NewRange()
|
||||
p.rn2[i] = object.NewRange()
|
||||
}
|
||||
|
||||
notches := c.splitPayload(p.id)
|
||||
|
||||
{ // node 1
|
||||
// [0:n2]
|
||||
p.rn1[0].SetLength(notches[1])
|
||||
|
||||
// [n2:n3]
|
||||
p.rn1[1].SetOffset(notches[1])
|
||||
p.rn1[1].SetLength(notches[2] - notches[1])
|
||||
|
||||
// [n3:full]
|
||||
p.rn1[2].SetOffset(notches[2])
|
||||
p.rn1[2].SetLength(notches[3] - notches[2])
|
||||
}
|
||||
|
||||
{ // node 2
|
||||
// [0:n1]
|
||||
p.rn2[0].SetLength(notches[0])
|
||||
|
||||
// [n1:n2]
|
||||
p.rn2[1].SetOffset(notches[0])
|
||||
p.rn2[1].SetLength(notches[1] - notches[0])
|
||||
|
||||
// [n2:full]
|
||||
p.rn2[2].SetOffset(notches[1])
|
||||
p.rn2[2].SetLength(notches[3] - notches[1])
|
||||
}
|
||||
}
|
||||
|
||||
func (c *Context) splitPayload(id oid.ID) []uint64 {
|
||||
var (
|
||||
prev uint64
|
||||
size = c.objectSize(id)
|
||||
notches = make([]uint64, 0, hashRangeNumber)
|
||||
)
|
||||
|
||||
for i := uint64(0); i < hashRangeNumber; i++ {
|
||||
if i < hashRangeNumber-1 {
|
||||
max := size - prev - (hashRangeNumber - i)
|
||||
if max == 0 {
|
||||
prev++
|
||||
} else {
|
||||
prev += rand.Uint64()%max + 1
|
||||
}
|
||||
} else {
|
||||
prev = size
|
||||
}
|
||||
|
||||
notches = append(notches, prev)
|
||||
}
|
||||
|
||||
return notches
|
||||
}
|
||||
|
||||
func (c *Context) collectHashes(ctx context.Context, p *gamePair) {
|
||||
fn := func(n netmap.NodeInfo, rngs []*object.Range) [][]byte {
|
||||
// Here we randomize the order a bit: the hypothesis is that this
|
||||
// makes it harder for an unscrupulous node to come up with a
|
||||
// reliable cheating strategy.
|
||||
order := make([]int, len(rngs))
|
||||
for i := range order {
|
||||
order[i] = i
|
||||
}
|
||||
rand.Shuffle(len(order), func(i, j int) { order[i], order[j] = order[j], order[i] })
|
||||
|
||||
var getRangeHashPrm GetRangeHashPrm
|
||||
getRangeHashPrm.CID = c.task.ContainerID()
|
||||
getRangeHashPrm.OID = p.id
|
||||
getRangeHashPrm.Node = n
|
||||
|
||||
res := make([][]byte, len(rngs))
|
||||
for _, i := range order {
|
||||
var sleepDur time.Duration
|
||||
if c.maxPDPSleep > 0 {
|
||||
sleepDur = time.Duration(rand.Uint64() % c.maxPDPSleep)
|
||||
}
|
||||
|
||||
c.log.Debug(logs.AuditorSleepBeforeGetRangeHash,
|
||||
zap.Stringer("interval", sleepDur),
|
||||
)
|
||||
|
||||
time.Sleep(sleepDur)
|
||||
|
||||
getRangeHashPrm.Range = rngs[i]
|
||||
|
||||
h, err := c.cnrCom.GetRangeHash(ctx, getRangeHashPrm)
|
||||
if err != nil {
|
||||
c.log.Debug(logs.AuditorCouldNotGetPayloadRangeHash,
|
||||
zap.Stringer("id", p.id),
|
||||
zap.String("node", netmap.StringifyPublicKey(n)),
|
||||
zap.String("error", err.Error()),
|
||||
)
|
||||
return res
|
||||
}
|
||||
res[i] = h
|
||||
}
|
||||
return res
|
||||
}
|
||||
|
||||
p.hh1 = fn(p.n1, p.rn1)
|
||||
p.hh2 = fn(p.n2, p.rn2)
|
||||
}
|
||||
|
||||
func (c *Context) analyzeHashes(p *gamePair) {
|
||||
if len(p.hh1) != hashRangeNumber-1 || len(p.hh2) != hashRangeNumber-1 {
|
||||
c.failNodesPDP(p.n1, p.n2)
|
||||
return
|
||||
}
|
||||
|
||||
h1, err := tz.Concat([][]byte{p.hh2[0], p.hh2[1]})
|
||||
if err != nil || !bytes.Equal(p.hh1[0], h1) {
|
||||
c.failNodesPDP(p.n1, p.n2)
|
||||
return
|
||||
}
|
||||
|
||||
h2, err := tz.Concat([][]byte{p.hh1[1], p.hh1[2]})
|
||||
if err != nil || !bytes.Equal(p.hh2[2], h2) {
|
||||
c.failNodesPDP(p.n1, p.n2)
|
||||
return
|
||||
}
|
||||
|
||||
fh, err := tz.Concat([][]byte{h1, h2})
|
||||
if err != nil || !bytes.Equal(fh, c.objectHomoHash(p.id)) {
|
||||
c.failNodesPDP(p.n1, p.n2)
|
||||
return
|
||||
}
|
||||
|
||||
c.passNodesPDP(p.n1, p.n2)
|
||||
}
|
||||
|
||||
func (c *Context) failNodesPDP(ns ...netmap.NodeInfo) {
|
||||
c.pairedMtx.Lock()
|
||||
|
||||
for i := range ns {
|
||||
c.pairedNodes[ns[i].Hash()].failedPDP = true
|
||||
}
|
||||
|
||||
c.pairedMtx.Unlock()
|
||||
}
|
||||
|
||||
func (c *Context) passNodesPDP(ns ...netmap.NodeInfo) {
|
||||
c.pairedMtx.Lock()
|
||||
|
||||
for i := range ns {
|
||||
c.pairedNodes[ns[i].Hash()].passedPDP = true
|
||||
}
|
||||
|
||||
c.pairedMtx.Unlock()
|
||||
}
|
||||
|
||||
func (c *Context) writePairsResult() {
|
||||
var failCount, okCount int
|
||||
|
||||
c.iteratePairedNodes(
|
||||
func(netmap.NodeInfo) { failCount++ },
|
||||
func(netmap.NodeInfo) { okCount++ },
|
||||
)
|
||||
|
||||
failedNodes := make([][]byte, 0, failCount)
|
||||
passedNodes := make([][]byte, 0, okCount)
|
||||
|
||||
c.iteratePairedNodes(
|
||||
func(n netmap.NodeInfo) {
|
||||
failedNodes = append(failedNodes, n.PublicKey())
|
||||
},
|
||||
func(n netmap.NodeInfo) {
|
||||
passedNodes = append(passedNodes, n.PublicKey())
|
||||
},
|
||||
)
|
||||
|
||||
c.report.SetPDPResults(passedNodes, failedNodes)
|
||||
}
|
||||
|
||||
func (c *Context) iteratePairedNodes(onFail, onPass func(netmap.NodeInfo)) {
|
||||
for _, pairedNode := range c.pairedNodes {
|
||||
if pairedNode.failedPDP {
|
||||
onFail(pairedNode.node)
|
||||
}
|
||||
|
||||
if pairedNode.passedPDP {
|
||||
onPass(pairedNode.node)
|
||||
}
|
||||
}
|
||||
}
|
|
@ -1,187 +0,0 @@
|
|||
package auditor
|
||||
|
||||
import (
|
||||
"context"
|
||||
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/internal/logs"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-sdk-go/netmap"
|
||||
oid "git.frostfs.info/TrueCloudLab/frostfs-sdk-go/object/id"
|
||||
"git.frostfs.info/TrueCloudLab/tzhash/tz"
|
||||
"go.uber.org/zap"
|
||||
)
|
||||
|
||||
const (
|
||||
hashRangeNumber = 4
|
||||
minGamePayloadSize = hashRangeNumber * tz.Size
|
||||
)
|
||||
|
||||
func (c *Context) executePoP(ctx context.Context) {
|
||||
c.buildCoverage(ctx)
|
||||
|
||||
c.report.SetPlacementCounters(
|
||||
c.counters.hit,
|
||||
c.counters.miss,
|
||||
c.counters.fail,
|
||||
)
|
||||
}
|
||||
|
||||
func (c *Context) buildCoverage(ctx context.Context) {
|
||||
policy := c.task.ContainerStructure().PlacementPolicy()
|
||||
|
||||
// select random member from another storage group
|
||||
// and process all placement vectors
|
||||
c.iterateSGMembersPlacementRand(func(id oid.ID, ind int, nodes []netmap.NodeInfo) bool {
|
||||
c.processObjectPlacement(ctx, id, nodes, policy.ReplicaNumberByIndex(ind))
|
||||
return c.containerCovered()
|
||||
})
|
||||
}
|
||||
|
||||
func (c *Context) containerCovered() bool {
|
||||
// number of container nodes can be calculated once
|
||||
return c.cnrNodesNum <= len(c.pairedNodes)
|
||||
}
|
||||
|
||||
func (c *Context) processObjectPlacement(ctx context.Context, id oid.ID, nodes []netmap.NodeInfo, replicas uint32) {
|
||||
var (
|
||||
ok uint32
|
||||
optimal bool
|
||||
|
||||
unpairedCandidate1, unpairedCandidate2 = -1, -1
|
||||
|
||||
pairedCandidate = -1
|
||||
)
|
||||
|
||||
var getHeaderPrm GetHeaderPrm
|
||||
getHeaderPrm.OID = id
|
||||
getHeaderPrm.CID = c.task.ContainerID()
|
||||
getHeaderPrm.NodeIsRelay = false
|
||||
|
||||
for i := 0; ok < replicas && i < len(nodes); i++ {
|
||||
getHeaderPrm.Node = nodes[i]
|
||||
|
||||
// try to get object header from node
|
||||
hdr, err := c.cnrCom.GetHeader(ctx, getHeaderPrm)
|
||||
if err != nil {
|
||||
c.log.Debug(logs.AuditorCouldNotGetObjectHeaderFromCandidate,
|
||||
zap.Stringer("id", id),
|
||||
zap.String("error", err.Error()),
|
||||
)
|
||||
|
||||
continue
|
||||
}
|
||||
|
||||
c.updateHeadResponses(hdr)
|
||||
|
||||
// increment success counter
|
||||
ok++
|
||||
|
||||
// update optimal flag
|
||||
optimal = ok == replicas && uint32(i) < replicas
|
||||
|
||||
// exclude small objects from coverage
|
||||
if c.objectSize(id) < minGamePayloadSize {
|
||||
continue
|
||||
}
|
||||
|
||||
// update potential candidates to be paired
|
||||
if _, ok := c.pairedNodes[nodes[i].Hash()]; !ok {
|
||||
if unpairedCandidate1 < 0 {
|
||||
unpairedCandidate1 = i
|
||||
} else if unpairedCandidate2 < 0 {
|
||||
unpairedCandidate2 = i
|
||||
}
|
||||
} else if pairedCandidate < 0 {
|
||||
pairedCandidate = i
|
||||
}
|
||||
}
|
||||
|
||||
if optimal {
|
||||
c.counters.hit++
|
||||
} else if ok == replicas {
|
||||
c.counters.miss++
|
||||
} else {
|
||||
c.counters.fail++
|
||||
}
|
||||
|
||||
if unpairedCandidate1 >= 0 {
|
||||
if unpairedCandidate2 >= 0 {
|
||||
c.composePair(id, nodes[unpairedCandidate1], nodes[unpairedCandidate2])
|
||||
} else if pairedCandidate >= 0 {
|
||||
c.composePair(id, nodes[unpairedCandidate1], nodes[pairedCandidate])
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (c *Context) composePair(id oid.ID, n1, n2 netmap.NodeInfo) {
|
||||
c.pairs = append(c.pairs, gamePair{
|
||||
n1: n1,
|
||||
n2: n2,
|
||||
id: id,
|
||||
})
|
||||
|
||||
c.pairedNodes[n1.Hash()] = &pairMemberInfo{
|
||||
node: n1,
|
||||
}
|
||||
c.pairedNodes[n2.Hash()] = &pairMemberInfo{
|
||||
node: n2,
|
||||
}
|
||||
}
|
||||
|
||||
func (c *Context) iterateSGMembersPlacementRand(f func(oid.ID, int, []netmap.NodeInfo) bool) {
|
||||
// iterate over storage groups members for all storage groups (one by one)
|
||||
// with randomly shuffled members
|
||||
c.iterateSGMembersRand(func(id oid.ID) bool {
|
||||
// build placement vector for the current object
|
||||
nn, err := c.buildPlacement(id)
|
||||
if err != nil {
|
||||
c.log.Debug(logs.AuditorCouldNotBuildPlacementForObject,
|
||||
zap.Stringer("id", id),
|
||||
zap.String("error", err.Error()),
|
||||
)
|
||||
|
||||
return false
|
||||
}
|
||||
|
||||
for i, nodes := range nn {
|
||||
if f(id, i, nodes) {
|
||||
return true
|
||||
}
|
||||
}
|
||||
|
||||
return false
|
||||
})
|
||||
}
|
||||
|
||||
func (c *Context) iterateSGMembersRand(f func(oid.ID) bool) {
|
||||
c.iterateSGInfo(func(members []oid.ID) bool {
|
||||
ln := len(members)
|
||||
|
||||
processed := make(map[uint64]struct{}, ln-1)
|
||||
|
||||
for len(processed) < ln {
|
||||
ind := nextRandUint64(uint64(ln), processed)
|
||||
processed[ind] = struct{}{}
|
||||
|
||||
if f(members[ind]) {
|
||||
return true
|
||||
}
|
||||
}
|
||||
|
||||
return false
|
||||
})
|
||||
}
|
||||
|
||||
func (c *Context) iterateSGInfo(f func([]oid.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
|
||||
// with unpredictable order
|
||||
for i := range c.sgMembersCache {
|
||||
if f(c.sgMembersCache[i]) {
|
||||
return
|
||||
}
|
||||
}
|
||||
}
|
|
@ -1,156 +0,0 @@
|
|||
package auditor
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"context"
|
||||
"sync"
|
||||
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/internal/logs"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/services/object_manager/placement"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/util/rand"
|
||||
containerSDK "git.frostfs.info/TrueCloudLab/frostfs-sdk-go/container"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-sdk-go/netmap"
|
||||
oid "git.frostfs.info/TrueCloudLab/frostfs-sdk-go/object/id"
|
||||
storagegroupSDK "git.frostfs.info/TrueCloudLab/frostfs-sdk-go/storagegroup"
|
||||
"git.frostfs.info/TrueCloudLab/tzhash/tz"
|
||||
"go.uber.org/zap"
|
||||
)
|
||||
|
||||
func (c *Context) executePoR(ctx context.Context) {
|
||||
wg := new(sync.WaitGroup)
|
||||
sgs := c.task.StorageGroupList()
|
||||
|
||||
for _, sg := range sgs {
|
||||
wg.Add(1)
|
||||
|
||||
if err := c.porWorkerPool.Submit(func() {
|
||||
c.checkStorageGroupPoR(ctx, sg.ID(), sg.StorageGroup())
|
||||
wg.Done()
|
||||
}); err != nil {
|
||||
wg.Done()
|
||||
}
|
||||
}
|
||||
|
||||
wg.Wait()
|
||||
c.porWorkerPool.Release()
|
||||
|
||||
c.report.SetPoRCounters(c.porRequests.Load(), c.porRetries.Load())
|
||||
}
|
||||
|
||||
func (c *Context) checkStorageGroupPoR(ctx context.Context, sgID oid.ID, sg storagegroupSDK.StorageGroup) {
|
||||
members := sg.Members()
|
||||
c.updateSGInfo(sgID, members)
|
||||
|
||||
var (
|
||||
tzHash []byte
|
||||
totalSize uint64
|
||||
|
||||
accRequests, accRetries uint32
|
||||
)
|
||||
|
||||
var getHeaderPrm GetHeaderPrm
|
||||
getHeaderPrm.CID = c.task.ContainerID()
|
||||
getHeaderPrm.NodeIsRelay = true
|
||||
|
||||
homomorphicHashingEnabled := !containerSDK.IsHomomorphicHashingDisabled(c.task.ContainerStructure())
|
||||
|
||||
for i := range members {
|
||||
flat, ok := c.getShuffledNodes(members[i], sgID)
|
||||
if !ok {
|
||||
continue
|
||||
}
|
||||
|
||||
getHeaderPrm.OID = members[i]
|
||||
|
||||
for j := range flat {
|
||||
accRequests++
|
||||
if j > 0 { // in best case audit get object header on first iteration
|
||||
accRetries++
|
||||
}
|
||||
|
||||
getHeaderPrm.Node = flat[j]
|
||||
|
||||
hdr, err := c.cnrCom.GetHeader(ctx, getHeaderPrm)
|
||||
if err != nil {
|
||||
c.log.Debug(logs.AuditorCantHeadObject,
|
||||
zap.String("remote_node", netmap.StringifyPublicKey(flat[j])),
|
||||
zap.Stringer("oid", members[i]),
|
||||
)
|
||||
|
||||
continue
|
||||
}
|
||||
|
||||
// update cache for PoR and PDP audit checks
|
||||
c.updateHeadResponses(hdr)
|
||||
|
||||
if homomorphicHashingEnabled {
|
||||
cs, _ := hdr.PayloadHomomorphicHash()
|
||||
if len(tzHash) == 0 {
|
||||
tzHash = cs.Value()
|
||||
} else {
|
||||
tzHash, err = tz.Concat([][]byte{
|
||||
tzHash,
|
||||
cs.Value(),
|
||||
})
|
||||
if err != nil {
|
||||
c.log.Debug(logs.AuditorCantConcatenateTzHash,
|
||||
zap.String("oid", members[i].String()),
|
||||
zap.String("error", err.Error()))
|
||||
|
||||
break
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
totalSize += hdr.PayloadSize()
|
||||
|
||||
break
|
||||
}
|
||||
}
|
||||
|
||||
c.porRequests.Add(accRequests)
|
||||
c.porRetries.Add(accRetries)
|
||||
|
||||
sizeCheck := sg.ValidationDataSize() == totalSize
|
||||
cs, _ := sg.ValidationDataHash()
|
||||
tzCheck := !homomorphicHashingEnabled || bytes.Equal(tzHash, cs.Value())
|
||||
|
||||
c.writeCheckReport(sizeCheck, tzCheck, sgID, sg, totalSize)
|
||||
}
|
||||
|
||||
func (c *Context) writeCheckReport(sizeCheck, tzCheck bool, sgID oid.ID, sg storagegroupSDK.StorageGroup, totalSize uint64) {
|
||||
if sizeCheck && tzCheck {
|
||||
c.report.PassedPoR(sgID)
|
||||
} else {
|
||||
if !sizeCheck {
|
||||
c.log.Debug(logs.AuditorStorageGroupSizeCheckFailed,
|
||||
zap.Uint64("expected", sg.ValidationDataSize()),
|
||||
zap.Uint64("got", totalSize))
|
||||
}
|
||||
|
||||
if !tzCheck {
|
||||
c.log.Debug(logs.AuditorStorageGroupTzHashCheckFailed)
|
||||
}
|
||||
|
||||
c.report.FailedPoR(sgID)
|
||||
}
|
||||
}
|
||||
|
||||
func (c *Context) getShuffledNodes(member oid.ID, sgID oid.ID) ([]netmap.NodeInfo, bool) {
|
||||
objectPlacement, err := c.buildPlacement(member)
|
||||
if err != nil {
|
||||
c.log.Info(logs.AuditorCantBuildPlacementForStorageGroupMember,
|
||||
zap.Stringer("sg", sgID),
|
||||
zap.String("member_id", member.String()),
|
||||
)
|
||||
|
||||
return nil, false
|
||||
}
|
||||
|
||||
flat := placement.FlattenNodes(objectPlacement)
|
||||
|
||||
rand.Shuffle(len(flat), func(i, j int) {
|
||||
flat[i], flat[j] = flat[j], flat[i]
|
||||
})
|
||||
return flat, true
|
||||
}
|
|
@ -1,18 +0,0 @@
|
|||
package auditor
|
||||
|
||||
import (
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/util/rand"
|
||||
)
|
||||
|
||||
// nextRandUint64 returns random uint64 number [0; n) outside exclude map.
|
||||
// Panics if len(exclude) >= n.
|
||||
func nextRandUint64(n uint64, exclude map[uint64]struct{}) uint64 {
|
||||
ln := uint64(len(exclude))
|
||||
ind := rand.Uint64() % (n - ln)
|
||||
|
||||
for i := ind; ; i++ {
|
||||
if _, ok := exclude[i]; !ok {
|
||||
return i
|
||||
}
|
||||
}
|
||||
}
|
|
@ -1,89 +0,0 @@
|
|||
package audit
|
||||
|
||||
import (
|
||||
"sync"
|
||||
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-sdk-go/audit"
|
||||
cid "git.frostfs.info/TrueCloudLab/frostfs-sdk-go/container/id"
|
||||
oid "git.frostfs.info/TrueCloudLab/frostfs-sdk-go/object/id"
|
||||
)
|
||||
|
||||
// Report tracks the progress of auditing container data.
|
||||
type Report struct {
|
||||
mu sync.RWMutex
|
||||
res audit.Result
|
||||
}
|
||||
|
||||
// Reporter is an interface of the entity that records
|
||||
// the data audit report.
|
||||
type Reporter interface {
|
||||
WriteReport(r *Report) error
|
||||
}
|
||||
|
||||
// NewReport creates and returns blank Report instance.
|
||||
func NewReport(cnr cid.ID) *Report {
|
||||
var rep Report
|
||||
rep.res.ForContainer(cnr)
|
||||
|
||||
return &rep
|
||||
}
|
||||
|
||||
// Result forms the structure of the data audit result.
|
||||
func (r *Report) Result() *audit.Result {
|
||||
r.mu.RLock()
|
||||
defer r.mu.RUnlock()
|
||||
|
||||
return &r.res
|
||||
}
|
||||
|
||||
// Complete completes audit report.
|
||||
func (r *Report) Complete() {
|
||||
r.mu.Lock()
|
||||
defer r.mu.Unlock()
|
||||
|
||||
r.res.Complete()
|
||||
}
|
||||
|
||||
// PassedPoR updates list of passed storage groups.
|
||||
func (r *Report) PassedPoR(sg oid.ID) {
|
||||
r.mu.Lock()
|
||||
defer r.mu.Unlock()
|
||||
|
||||
r.res.SubmitPassedStorageGroup(sg)
|
||||
}
|
||||
|
||||
// FailedPoR updates list of failed storage groups.
|
||||
func (r *Report) FailedPoR(sg oid.ID) {
|
||||
r.mu.Lock()
|
||||
defer r.mu.Unlock()
|
||||
|
||||
r.res.SubmitFailedStorageGroup(sg)
|
||||
}
|
||||
|
||||
// SetPlacementCounters sets counters of compliance with placement.
|
||||
func (r *Report) SetPlacementCounters(hit, miss, fail uint32) {
|
||||
r.mu.Lock()
|
||||
defer r.mu.Unlock()
|
||||
|
||||
r.res.SetHits(hit)
|
||||
r.res.SetMisses(miss)
|
||||
r.res.SetFailures(fail)
|
||||
}
|
||||
|
||||
// SetPDPResults sets lists of nodes according to their PDP results.
|
||||
func (r *Report) SetPDPResults(passed, failed [][]byte) {
|
||||
r.mu.Lock()
|
||||
defer r.mu.Unlock()
|
||||
|
||||
r.res.SubmitPassedStorageNodes(passed)
|
||||
r.res.SubmitFailedStorageNodes(failed)
|
||||
}
|
||||
|
||||
// SetPoRCounters sets amounts of head requests and retries at PoR audit stage.
|
||||
func (r *Report) SetPoRCounters(requests, retries uint32) {
|
||||
r.mu.Lock()
|
||||
defer r.mu.Unlock()
|
||||
|
||||
r.res.SetRequestsPoR(requests)
|
||||
r.res.SetRetriesPoR(retries)
|
||||
}
|
|
@ -1,120 +0,0 @@
|
|||
package audit
|
||||
|
||||
import (
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/core/storagegroup"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-sdk-go/container"
|
||||
cid "git.frostfs.info/TrueCloudLab/frostfs-sdk-go/container/id"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-sdk-go/netmap"
|
||||
)
|
||||
|
||||
// Task groups groups the container audit parameters.
|
||||
type Task struct {
|
||||
cancelCh <-chan struct{}
|
||||
|
||||
reporter Reporter
|
||||
|
||||
idCnr cid.ID
|
||||
|
||||
cnr container.Container
|
||||
|
||||
nm *netmap.NetMap
|
||||
|
||||
cnrNodes [][]netmap.NodeInfo
|
||||
|
||||
sgList []storagegroup.StorageGroup
|
||||
}
|
||||
|
||||
// WithReporter sets audit report writer.
|
||||
func (t *Task) WithReporter(r Reporter) *Task {
|
||||
if t != nil {
|
||||
t.reporter = r
|
||||
}
|
||||
|
||||
return t
|
||||
}
|
||||
|
||||
// Reporter returns audit report writer.
|
||||
func (t *Task) Reporter() Reporter {
|
||||
return t.reporter
|
||||
}
|
||||
|
||||
func (t *Task) WithCancelChannel(ch <-chan struct{}) *Task {
|
||||
if ch != nil {
|
||||
t.cancelCh = ch
|
||||
}
|
||||
return t
|
||||
}
|
||||
|
||||
func (t *Task) CancelChannel() <-chan struct{} {
|
||||
return t.cancelCh
|
||||
}
|
||||
|
||||
// WithContainerID sets identifier of the container under audit.
|
||||
func (t *Task) WithContainerID(cnr cid.ID) *Task {
|
||||
if t != nil {
|
||||
t.idCnr = cnr
|
||||
}
|
||||
|
||||
return t
|
||||
}
|
||||
|
||||
// ContainerID returns identifier of the container under audit.
|
||||
func (t *Task) ContainerID() cid.ID {
|
||||
return t.idCnr
|
||||
}
|
||||
|
||||
// WithContainerStructure sets structure of the container under audit.
|
||||
func (t *Task) WithContainerStructure(cnr container.Container) *Task {
|
||||
if t != nil {
|
||||
t.cnr = cnr
|
||||
}
|
||||
|
||||
return t
|
||||
}
|
||||
|
||||
// ContainerStructure returns structure of the container under audit.
|
||||
func (t *Task) ContainerStructure() container.Container {
|
||||
return t.cnr
|
||||
}
|
||||
|
||||
// WithContainerNodes sets nodes in the container under audit.
|
||||
func (t *Task) WithContainerNodes(cnrNodes [][]netmap.NodeInfo) *Task {
|
||||
if t != nil {
|
||||
t.cnrNodes = cnrNodes
|
||||
}
|
||||
|
||||
return t
|
||||
}
|
||||
|
||||
// NetworkMap returns network map of audit epoch.
|
||||
func (t *Task) NetworkMap() *netmap.NetMap {
|
||||
return t.nm
|
||||
}
|
||||
|
||||
// WithNetworkMap sets network map of audit epoch.
|
||||
func (t *Task) WithNetworkMap(nm *netmap.NetMap) *Task {
|
||||
if t != nil {
|
||||
t.nm = nm
|
||||
}
|
||||
|
||||
return t
|
||||
}
|
||||
|
||||
// ContainerNodes returns nodes in the container under audit.
|
||||
func (t *Task) ContainerNodes() [][]netmap.NodeInfo {
|
||||
return t.cnrNodes
|
||||
}
|
||||
|
||||
// WithStorageGroupList sets a list of storage groups from container under audit.
|
||||
func (t *Task) WithStorageGroupList(sgList []storagegroup.StorageGroup) *Task {
|
||||
if t != nil {
|
||||
t.sgList = sgList
|
||||
}
|
||||
|
||||
return t
|
||||
}
|
||||
|
||||
// StorageGroupList returns list of storage groups from container under audit.
|
||||
func (t *Task) StorageGroupList() []storagegroup.StorageGroup {
|
||||
return t.sgList
|
||||
}
|
|
@ -1,85 +0,0 @@
|
|||
package audittask
|
||||
|
||||
import (
|
||||
"context"
|
||||
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/internal/logs"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/services/audit"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/services/audit/auditor"
|
||||
"go.uber.org/zap"
|
||||
)
|
||||
|
||||
// Listen starts the process of processing tasks from the queue.
|
||||
//
|
||||
// The listener is terminated by context.
|
||||
func (m *Manager) Listen(ctx context.Context) {
|
||||
m.log.Info(logs.TaskmanagerProcessRoutine,
|
||||
zap.Uint32("queue_capacity", m.queueCap),
|
||||
)
|
||||
|
||||
m.ch = make(chan *audit.Task, m.queueCap)
|
||||
|
||||
for {
|
||||
select {
|
||||
case <-ctx.Done():
|
||||
m.log.Warn(logs.TaskmanagerStopListenerByContext,
|
||||
zap.String("error", ctx.Err().Error()),
|
||||
)
|
||||
m.workerPool.Release()
|
||||
|
||||
return
|
||||
case task, ok := <-m.ch:
|
||||
if !ok {
|
||||
m.log.Warn(logs.TaskmanagerQueueChannelIsClosed)
|
||||
return
|
||||
}
|
||||
|
||||
tCtx, tCancel := context.WithCancel(ctx) // cancel task in case of listen cancel
|
||||
go func() {
|
||||
select {
|
||||
case <-tCtx.Done(): // listen cancelled or task completed
|
||||
return
|
||||
case <-task.CancelChannel(): // new epoch
|
||||
tCancel()
|
||||
}
|
||||
}()
|
||||
|
||||
m.handleTask(tCtx, task, tCancel)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (m *Manager) handleTask(ctx context.Context, task *audit.Task, onCompleted func()) {
|
||||
pdpPool, err := m.pdpPoolGenerator()
|
||||
if err != nil {
|
||||
m.log.Error(logs.TaskmanagerCouldNotGeneratePDPWorkerPool,
|
||||
zap.String("error", err.Error()),
|
||||
)
|
||||
onCompleted()
|
||||
return
|
||||
}
|
||||
|
||||
porPool, err := m.pdpPoolGenerator()
|
||||
if err != nil {
|
||||
m.log.Error(logs.TaskmanagerCouldNotGeneratePoRWorkerPool,
|
||||
zap.String("error", err.Error()),
|
||||
)
|
||||
onCompleted()
|
||||
return
|
||||
}
|
||||
|
||||
auditContext := m.generateContext(task).
|
||||
WithPDPWorkerPool(pdpPool).
|
||||
WithPoRWorkerPool(porPool)
|
||||
|
||||
if err := m.workerPool.Submit(func() { auditContext.Execute(ctx, onCompleted) }); err != nil {
|
||||
// may be we should report it
|
||||
m.log.Warn(logs.TaskmanagerCouldNotSubmitAuditTask)
|
||||
onCompleted()
|
||||
}
|
||||
}
|
||||
|
||||
func (m *Manager) generateContext(task *audit.Task) *auditor.Context {
|
||||
return auditor.NewContext(m.ctxPrm).
|
||||
WithTask(task)
|
||||
}
|
|
@ -1,107 +0,0 @@
|
|||
package audittask
|
||||
|
||||
import (
|
||||
"time"
|
||||
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/services/audit"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/services/audit/auditor"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/util"
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/util/logger"
|
||||
"go.uber.org/zap"
|
||||
)
|
||||
|
||||
// Manager represents an entity performing data audit tasks.
|
||||
type Manager struct {
|
||||
*cfg
|
||||
|
||||
ch chan *audit.Task
|
||||
}
|
||||
|
||||
// Option is a Manager's constructor option.
|
||||
type Option func(*cfg)
|
||||
|
||||
type cfg struct {
|
||||
queueCap uint32
|
||||
|
||||
log *logger.Logger
|
||||
|
||||
ctxPrm auditor.ContextPrm
|
||||
|
||||
workerPool util.WorkerPool
|
||||
|
||||
pdpPoolGenerator, porPoolGenerator func() (util.WorkerPool, error)
|
||||
}
|
||||
|
||||
func defaultCfg() *cfg {
|
||||
return &cfg{
|
||||
log: &logger.Logger{Logger: zap.L()},
|
||||
}
|
||||
}
|
||||
|
||||
// New creates, initializes and returns new Manager instance.
|
||||
func New(opts ...Option) *Manager {
|
||||
c := defaultCfg()
|
||||
|
||||
for i := range opts {
|
||||
opts[i](c)
|
||||
}
|
||||
|
||||
return &Manager{
|
||||
cfg: c,
|
||||
}
|
||||
}
|
||||
|
||||
// WithLogger returns option to specify Manager's logger.
|
||||
func WithLogger(l *logger.Logger) Option {
|
||||
return func(c *cfg) {
|
||||
c.log = &logger.Logger{Logger: l.With(zap.String("component", "Audit task manager"))}
|
||||
c.ctxPrm.SetLogger(l)
|
||||
}
|
||||
}
|
||||
|
||||
// WithWorkerPool returns option to set worker pool
|
||||
// for task execution.
|
||||
func WithWorkerPool(p util.WorkerPool) Option {
|
||||
return func(c *cfg) {
|
||||
c.workerPool = p
|
||||
}
|
||||
}
|
||||
|
||||
// WithQueueCapacity returns option to set task queue capacity.
|
||||
func WithQueueCapacity(capacity uint32) Option {
|
||||
return func(c *cfg) {
|
||||
c.queueCap = capacity
|
||||
}
|
||||
}
|
||||
|
||||
// WithContainerCommunicator returns option to set component of communication
|
||||
// with container nodes.
|
||||
func WithContainerCommunicator(cnrCom auditor.ContainerCommunicator) Option {
|
||||
return func(c *cfg) {
|
||||
c.ctxPrm.SetContainerCommunicator(cnrCom)
|
||||
}
|
||||
}
|
||||
|
||||
// WithMaxPDPSleepInterval returns option to set maximum sleep interval
|
||||
// between range hash requests as part of PDP check.
|
||||
func WithMaxPDPSleepInterval(dur time.Duration) Option {
|
||||
return func(c *cfg) {
|
||||
c.ctxPrm.SetMaxPDPSleep(dur)
|
||||
}
|
||||
}
|
||||
|
||||
// WithPDPWorkerPoolGenerator returns option to set worker pool for PDP pairs processing.
|
||||
// Callback caller owns returned pool and must release it appropriately.
|
||||
func WithPDPWorkerPoolGenerator(f func() (util.WorkerPool, error)) Option {
|
||||
return func(c *cfg) {
|
||||
c.pdpPoolGenerator = f
|
||||
}
|
||||
}
|
||||
|
||||
// WithPoRWorkerPoolGenerator returns option to set worker pool for PoR SG processing.
|
||||
// Callback caller owns returned pool and must release it appropriately.
|
||||
func WithPoRWorkerPoolGenerator(f func() (util.WorkerPool, error)) Option {
|
||||
return func(c *cfg) {
|
||||
c.porPoolGenerator = f
|
||||
}
|
||||
}
|
|
@ -1,10 +0,0 @@
|
|||
package audittask
|
||||
|
||||
import (
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/services/audit"
|
||||
)
|
||||
|
||||
// PushTask adds a task to the queue for processing.
|
||||
func (m *Manager) PushTask(t *audit.Task) {
|
||||
m.ch <- t
|
||||
}
|
|
@ -1,11 +0,0 @@
|
|||
package audittask
|
||||
|
||||
// Reset pops all tasks from the queue.
|
||||
// Returns amount of popped elements.
|
||||
func (m *Manager) Reset() (popped int) {
|
||||
for ; len(m.ch) > 0; popped++ {
|
||||
<-m.ch
|
||||
}
|
||||
|
||||
return
|
||||
}
|
Loading…
Reference in a new issue