minio/internal/dsync/drwmutex.go

746 lines
20 KiB
Go

// Copyright (c) 2015-2021 MinIO, Inc.
//
// This file is part of MinIO Object Storage stack
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU Affero General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// This program is distributed in the hope that it will be useful
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Affero General Public License for more details.
//
// You should have received a copy of the GNU Affero General Public License
// along with this program. If not, see <http://www.gnu.org/licenses/>.
package dsync
import (
"context"
"errors"
"math/rand"
"sort"
"strconv"
"sync"
"time"
xioutil "github.com/minio/minio/internal/ioutil"
"github.com/minio/minio/internal/mcontext"
"github.com/minio/pkg/v3/console"
"github.com/minio/pkg/v3/env"
)
// Indicator if logging is enabled.
var dsyncLog bool
// Retry unit interval
var lockRetryMinInterval time.Duration
var lockRetryBackOff func(*rand.Rand, uint) time.Duration
func init() {
// Check for MINIO_DSYNC_TRACE env variable, if set logging will be enabled for failed REST operations.
dsyncLog = env.Get("_MINIO_DSYNC_TRACE", "0") == "1"
lockRetryMinInterval = 250 * time.Millisecond
if lri := env.Get("_MINIO_LOCK_RETRY_INTERVAL", ""); lri != "" {
v, err := strconv.Atoi(lri)
if err != nil {
panic(err)
}
lockRetryMinInterval = time.Duration(v) * time.Millisecond
}
lockRetryBackOff = backoffWait(
lockRetryMinInterval,
100*time.Millisecond,
5*time.Second,
)
}
func log(format string, data ...interface{}) {
if dsyncLog {
console.Printf(format, data...)
}
}
const (
// dRWMutexAcquireTimeout - default tolerance limit to wait for lock acquisition before.
drwMutexAcquireTimeout = 1 * time.Second // 1 second.
// dRWMutexRefreshTimeout - default timeout for the refresh call
drwMutexRefreshCallTimeout = 5 * time.Second
// dRWMutexUnlockTimeout - default timeout for the unlock call
drwMutexUnlockCallTimeout = 30 * time.Second
// dRWMutexForceUnlockTimeout - default timeout for the unlock call
drwMutexForceUnlockCallTimeout = 30 * time.Second
// dRWMutexRefreshInterval - default the interval between two refresh calls
drwMutexRefreshInterval = 10 * time.Second
drwMutexInfinite = 1<<63 - 1
)
// Timeouts are timeouts for specific operations.
type Timeouts struct {
// Acquire - tolerance limit to wait for lock acquisition before.
Acquire time.Duration
// RefreshCall - timeout for the refresh call
RefreshCall time.Duration
// UnlockCall - timeout for the unlock call
UnlockCall time.Duration
// ForceUnlockCall - timeout for the force unlock call
ForceUnlockCall time.Duration
}
// DefaultTimeouts contains default timeouts.
var DefaultTimeouts = Timeouts{
Acquire: drwMutexAcquireTimeout,
RefreshCall: drwMutexRefreshCallTimeout,
UnlockCall: drwMutexUnlockCallTimeout,
ForceUnlockCall: drwMutexForceUnlockCallTimeout,
}
// A DRWMutex is a distributed mutual exclusion lock.
type DRWMutex struct {
Names []string
writeLocks []string // Array of nodes that granted a write lock
readLocks []string // Array of array of nodes that granted reader locks
rng *rand.Rand
m sync.Mutex // Mutex to prevent multiple simultaneous locks from this node
clnt *Dsync
cancelRefresh context.CancelFunc
refreshInterval time.Duration
lockRetryMinInterval time.Duration
}
// Granted - represents a structure of a granted lock.
type Granted struct {
index int
lockUID string // Locked if set with UID string, unlocked if empty
}
func (g *Granted) isLocked() bool {
return isLocked(g.lockUID)
}
func isLocked(uid string) bool {
return len(uid) > 0
}
// NewDRWMutex - initializes a new dsync RW mutex.
func NewDRWMutex(clnt *Dsync, names ...string) *DRWMutex {
restClnts, _ := clnt.GetLockers()
sort.Strings(names)
return &DRWMutex{
writeLocks: make([]string, len(restClnts)),
readLocks: make([]string, len(restClnts)),
Names: names,
clnt: clnt,
rng: rand.New(&lockedRandSource{src: rand.NewSource(time.Now().UTC().UnixNano())}),
refreshInterval: drwMutexRefreshInterval,
lockRetryMinInterval: lockRetryMinInterval,
}
}
// Lock holds a write lock on dm.
//
// If the lock is already in use, the calling go routine
// blocks until the mutex is available.
func (dm *DRWMutex) Lock(id, source string) {
isReadLock := false
dm.lockBlocking(context.Background(), nil, id, source, isReadLock, Options{
Timeout: drwMutexInfinite,
})
}
// Options lock options.
type Options struct {
Timeout time.Duration
RetryInterval time.Duration
}
// GetLock tries to get a write lock on dm before the timeout elapses.
//
// If the lock is already in use, the calling go routine
// blocks until either the mutex becomes available and return success or
// more time has passed than the timeout value and return false.
func (dm *DRWMutex) GetLock(ctx context.Context, cancel context.CancelFunc, id, source string, opts Options) (locked bool) {
isReadLock := false
return dm.lockBlocking(ctx, cancel, id, source, isReadLock, opts)
}
// RLock holds a read lock on dm.
//
// If one or more read locks are already in use, it will grant another lock.
// Otherwise the calling go routine blocks until the mutex is available.
func (dm *DRWMutex) RLock(id, source string) {
isReadLock := true
dm.lockBlocking(context.Background(), nil, id, source, isReadLock, Options{
Timeout: drwMutexInfinite,
})
}
// GetRLock tries to get a read lock on dm before the timeout elapses.
//
// If one or more read locks are already in use, it will grant another lock.
// Otherwise the calling go routine blocks until either the mutex becomes
// available and return success or more time has passed than the timeout
// value and return false.
func (dm *DRWMutex) GetRLock(ctx context.Context, cancel context.CancelFunc, id, source string, opts Options) (locked bool) {
isReadLock := true
return dm.lockBlocking(ctx, cancel, id, source, isReadLock, opts)
}
// lockBlocking will try to acquire either a read or a write lock
//
// The function will loop using a built-in timing randomized back-off
// algorithm until either the lock is acquired successfully or more
// time has elapsed than the timeout value.
func (dm *DRWMutex) lockBlocking(ctx context.Context, lockLossCallback func(), id, source string, isReadLock bool, opts Options) (locked bool) {
restClnts, _ := dm.clnt.GetLockers()
// Create lock array to capture the successful lockers
locks := make([]string, len(restClnts))
// Add total timeout
ctx, cancel := context.WithTimeout(ctx, opts.Timeout)
defer cancel()
// Tolerance is not set, defaults to half of the locker clients.
tolerance := len(restClnts) / 2
// Quorum is effectively = total clients subtracted with tolerance limit
quorum := len(restClnts) - tolerance
if !isReadLock {
// In situations for write locks, as a special case
// to avoid split brains we make sure to acquire
// quorum + 1 when tolerance is exactly half of the
// total locker clients.
if quorum == tolerance {
quorum++
}
}
log("lockBlocking %s/%s for %#v: lockType readLock(%t), additional opts: %#v, quorum: %d, tolerance: %d, lockClients: %d\n", id, source, dm.Names, isReadLock, opts, quorum, tolerance, len(restClnts))
tolerance = len(restClnts) - quorum
attempt := uint(0)
for {
select {
case <-ctx.Done():
return false
default:
// Try to acquire the lock.
if locked = lock(ctx, dm.clnt, &locks, id, source, isReadLock, tolerance, quorum, dm.Names...); locked {
dm.m.Lock()
// If success, copy array to object
if isReadLock {
copy(dm.readLocks, locks)
} else {
copy(dm.writeLocks, locks)
}
dm.m.Unlock()
log("lockBlocking %s/%s for %#v: granted\n", id, source, dm.Names)
// Refresh lock continuously and cancel if there is no quorum in the lock anymore
dm.startContinuousLockRefresh(lockLossCallback, id, source, quorum)
return locked
}
switch {
case opts.RetryInterval < 0:
return false
case opts.RetryInterval > 0:
time.Sleep(opts.RetryInterval)
default:
attempt++
time.Sleep(lockRetryBackOff(dm.rng, attempt))
}
}
}
}
func (dm *DRWMutex) startContinuousLockRefresh(lockLossCallback func(), id, source string, quorum int) {
ctx, cancel := context.WithCancel(context.Background())
dm.m.Lock()
dm.cancelRefresh = cancel
dm.m.Unlock()
go func() {
defer cancel()
refreshTimer := time.NewTimer(dm.refreshInterval)
defer refreshTimer.Stop()
for {
select {
case <-ctx.Done():
return
case <-refreshTimer.C:
noQuorum, err := refreshLock(ctx, dm.clnt, id, source, quorum)
if err == nil && noQuorum {
// Clean the lock locally and in remote nodes
forceUnlock(ctx, dm.clnt, id)
// Execute the caller lock loss callback
if lockLossCallback != nil {
lockLossCallback()
}
return
}
refreshTimer.Reset(dm.refreshInterval)
}
}
}()
}
func forceUnlock(ctx context.Context, ds *Dsync, id string) {
ctx, cancel := context.WithTimeout(ctx, ds.Timeouts.ForceUnlockCall)
defer cancel()
restClnts, _ := ds.GetLockers()
args := LockArgs{
UID: id,
}
var wg sync.WaitGroup
for index, c := range restClnts {
wg.Add(1)
// Send refresh request to all nodes
go func(index int, c NetLocker) {
defer wg.Done()
c.ForceUnlock(ctx, args)
}(index, c)
}
wg.Wait()
}
type refreshResult struct {
offline bool
refreshed bool
}
// Refresh the given lock in all nodes, return true to indicate if a lock
// does not exist in enough quorum nodes.
func refreshLock(ctx context.Context, ds *Dsync, id, source string, quorum int) (bool, error) {
restClnts, _ := ds.GetLockers()
// Create buffered channel of size equal to total number of nodes.
ch := make(chan refreshResult, len(restClnts))
var wg sync.WaitGroup
args := LockArgs{
UID: id,
}
for index, c := range restClnts {
wg.Add(1)
// Send refresh request to all nodes
go func(index int, c NetLocker) {
defer wg.Done()
if c == nil {
ch <- refreshResult{offline: true}
return
}
ctx, cancel := context.WithTimeout(ctx, ds.Timeouts.RefreshCall)
defer cancel()
refreshed, err := c.Refresh(ctx, args)
if err != nil {
ch <- refreshResult{offline: true}
log("dsync: Unable to call Refresh failed with %s for %#v at %s\n", err, args, c)
} else {
ch <- refreshResult{refreshed: refreshed}
log("dsync: Refresh returned false for %#v at %s\n", args, c)
}
}(index, c)
}
// Wait until we have either
//
// a) received all refresh responses
// b) received too many refreshed for quorum to be still possible
// c) timed out
//
lockNotFound, lockRefreshed := 0, 0
done := false
for i := 0; i < len(restClnts); i++ {
select {
case refreshResult := <-ch:
if refreshResult.offline {
continue
}
if refreshResult.refreshed {
lockRefreshed++
} else {
lockNotFound++
}
if lockRefreshed >= quorum || lockNotFound > len(restClnts)-quorum {
done = true
}
case <-ctx.Done():
// Refreshing is canceled
return false, ctx.Err()
}
if done {
break
}
}
// We may have some unused results in ch, release them async.
go func() {
wg.Wait()
xioutil.SafeClose(ch)
for range ch {
}
}()
noQuorum := lockNotFound > len(restClnts)-quorum
return noQuorum, nil
}
// lock tries to acquire the distributed lock, returning true or false.
func lock(ctx context.Context, ds *Dsync, locks *[]string, id, source string, isReadLock bool, tolerance, quorum int, names ...string) bool {
for i := range *locks {
(*locks)[i] = ""
}
restClnts, owner := ds.GetLockers()
// Create buffered channel of size equal to total number of nodes.
ch := make(chan Granted, len(restClnts))
var wg sync.WaitGroup
args := LockArgs{
Owner: owner,
UID: id,
Resources: names,
Source: source,
Quorum: &quorum,
}
// Combined timeout for the lock attempt.
ctx, cancel := context.WithTimeout(ctx, ds.Timeouts.Acquire)
defer cancel()
// Special context for NetLockers - do not use timeouts.
// Also, pass the trace context info if found for debugging
netLockCtx := context.Background()
tc, ok := ctx.Value(mcontext.ContextTraceKey).(*mcontext.TraceCtxt)
if ok {
netLockCtx = context.WithValue(netLockCtx, mcontext.ContextTraceKey, tc)
}
for index, c := range restClnts {
wg.Add(1)
// broadcast lock request to all nodes
go func(index int, isReadLock bool, c NetLocker) {
defer wg.Done()
g := Granted{index: index}
if c == nil {
log("dsync: nil locker\n")
ch <- g
return
}
var locked bool
var err error
if isReadLock {
if locked, err = c.RLock(netLockCtx, args); err != nil {
log("dsync: Unable to call RLock failed with %s for %#v at %s\n", err, args, c)
}
} else {
if locked, err = c.Lock(netLockCtx, args); err != nil {
log("dsync: Unable to call Lock failed with %s for %#v at %s\n", err, args, c)
}
}
if locked {
g.lockUID = args.UID
}
ch <- g
}(index, isReadLock, c)
}
// Wait until we have either
//
// a) received all lock responses
// b) received too many 'non-'locks for quorum to be still possible
// c) timed out
//
i, locksFailed := 0, 0
done := false
for ; i < len(restClnts); i++ { // Loop until we acquired all locks
select {
case grant := <-ch:
if grant.isLocked() {
// Mark that this node has acquired the lock
(*locks)[grant.index] = grant.lockUID
} else {
locksFailed++
if locksFailed > tolerance {
// We know that we are not going to get the lock anymore,
// so exit out and release any locks that did get acquired
done = true
}
}
case <-ctx.Done():
// Capture timedout locks as failed or took too long
locksFailed++
if locksFailed > tolerance {
// We know that we are not going to get the lock anymore,
// so exit out and release any locks that did get acquired
done = true
}
}
if done {
break
}
}
quorumLocked := checkQuorumLocked(locks, quorum) && locksFailed <= tolerance
if !quorumLocked {
log("dsync: Unable to acquire lock in quorum %#v\n", args)
// Release all acquired locks without quorum.
if !releaseAll(ctx, ds, tolerance, owner, locks, isReadLock, restClnts, names...) {
log("Unable to release acquired locks, these locks will expire automatically %#v\n", args)
}
}
// We may have some unused results in ch, release them async.
go func() {
wg.Wait()
xioutil.SafeClose(ch)
for grantToBeReleased := range ch {
if grantToBeReleased.isLocked() {
// release abandoned lock
log("Releasing abandoned lock\n")
sendRelease(ctx, ds, restClnts[grantToBeReleased.index],
owner, grantToBeReleased.lockUID, isReadLock, names...)
}
}
}()
return quorumLocked
}
// checkFailedUnlocks determines whether we have sufficiently unlocked all
// resources to ensure no deadlocks for future callers
func checkFailedUnlocks(locks []string, tolerance int) bool {
unlocksFailed := 0
for lockID := range locks {
if isLocked(locks[lockID]) {
unlocksFailed++
}
}
// Unlock failures are higher than tolerance limit
// for this instance of unlocker, we should let the
// caller know that lock is not successfully released
// yet.
if len(locks)-tolerance == tolerance {
// In case of split brain scenarios where
// tolerance is exactly half of the len(*locks)
// then we need to make sure we have unlocked
// upto tolerance+1 - especially for RUnlock
// to ensure that we don't end up with active
// read locks on the resource after unlocking
// only half of the lockers.
return unlocksFailed >= tolerance
}
return unlocksFailed > tolerance
}
// checkQuorumLocked determines whether we have locked the required quorum of underlying locks or not
func checkQuorumLocked(locks *[]string, quorum int) bool {
count := 0
for _, uid := range *locks {
if isLocked(uid) {
count++
}
}
return count >= quorum
}
// releaseAll releases all locks that are marked as locked
func releaseAll(ctx context.Context, ds *Dsync, tolerance int, owner string, locks *[]string, isReadLock bool, restClnts []NetLocker, names ...string) bool {
var wg sync.WaitGroup
for lockID := range restClnts {
wg.Add(1)
go func(lockID int) {
defer wg.Done()
if sendRelease(ctx, ds, restClnts[lockID], owner, (*locks)[lockID], isReadLock, names...) {
(*locks)[lockID] = ""
}
}(lockID)
}
wg.Wait()
// Return true if releaseAll was successful, otherwise we return 'false'
// to indicate we haven't sufficiently unlocked lockers to avoid deadlocks.
//
// Caller may use this as an indication to call again.
return !checkFailedUnlocks(*locks, tolerance)
}
// Unlock unlocks the write lock.
//
// It is a run-time error if dm is not locked on entry to Unlock.
func (dm *DRWMutex) Unlock(ctx context.Context) {
dm.m.Lock()
dm.cancelRefresh()
dm.m.Unlock()
restClnts, owner := dm.clnt.GetLockers()
// create temp array on stack
locks := make([]string, len(restClnts))
{
dm.m.Lock()
defer dm.m.Unlock()
// Check if minimally a single bool is set in the writeLocks array
lockFound := false
for _, uid := range dm.writeLocks {
if isLocked(uid) {
lockFound = true
break
}
}
if !lockFound {
panic("Trying to Unlock() while no Lock() is active")
}
// Copy write locks to stack array
copy(locks, dm.writeLocks)
}
// Tolerance is not set, defaults to half of the locker clients.
tolerance := len(restClnts) / 2
isReadLock := false
started := time.Now()
// Do async unlocking.
// This means unlock will no longer block on the network or missing quorum.
go func() {
ctx, done := context.WithTimeout(ctx, drwMutexUnlockCallTimeout)
defer done()
for !releaseAll(ctx, dm.clnt, tolerance, owner, &locks, isReadLock, restClnts, dm.Names...) {
time.Sleep(time.Duration(dm.rng.Float64() * float64(dm.lockRetryMinInterval)))
if time.Since(started) > dm.clnt.Timeouts.UnlockCall {
return
}
}
}()
}
// RUnlock releases a read lock held on dm.
//
// It is a run-time error if dm is not locked on entry to RUnlock.
func (dm *DRWMutex) RUnlock(ctx context.Context) {
dm.m.Lock()
dm.cancelRefresh()
dm.m.Unlock()
restClnts, owner := dm.clnt.GetLockers()
// create temp array on stack
locks := make([]string, len(restClnts))
{
dm.m.Lock()
defer dm.m.Unlock()
// Check if minimally a single bool is set in the writeLocks array
lockFound := false
for _, uid := range dm.readLocks {
if isLocked(uid) {
lockFound = true
break
}
}
if !lockFound {
panic("Trying to RUnlock() while no RLock() is active")
}
// Copy write locks to stack array
copy(locks, dm.readLocks)
}
// Tolerance is not set, defaults to half of the locker clients.
tolerance := len(restClnts) / 2
isReadLock := true
started := time.Now()
// Do async unlocking.
// This means unlock will no longer block on the network or missing quorum.
go func() {
for !releaseAll(ctx, dm.clnt, tolerance, owner, &locks, isReadLock, restClnts, dm.Names...) {
time.Sleep(time.Duration(dm.rng.Float64() * float64(dm.lockRetryMinInterval)))
// If we have been waiting for more than the force unlock timeout, return
// Remotes will have canceled due to the missing refreshes anyway.
if time.Since(started) > dm.clnt.Timeouts.UnlockCall {
return
}
}
}()
}
// sendRelease sends a release message to a node that previously granted a lock
func sendRelease(ctx context.Context, ds *Dsync, c NetLocker, owner string, uid string, isReadLock bool, names ...string) bool {
if c == nil {
log("Unable to call RUnlock failed with %s\n", errors.New("netLocker is offline"))
return false
}
if len(uid) == 0 {
return false
}
args := LockArgs{
Owner: owner,
UID: uid,
Resources: names,
}
netLockCtx, cancel := context.WithTimeout(context.Background(), ds.Timeouts.UnlockCall)
defer cancel()
tc, ok := ctx.Value(mcontext.ContextTraceKey).(*mcontext.TraceCtxt)
if ok {
netLockCtx = context.WithValue(netLockCtx, mcontext.ContextTraceKey, tc)
}
if isReadLock {
if _, err := c.RUnlock(netLockCtx, args); err != nil {
log("dsync: Unable to call RUnlock failed with %s for %#v at %s\n", err, args, c)
return false
}
} else {
if _, err := c.Unlock(netLockCtx, args); err != nil {
log("dsync: Unable to call Unlock failed with %s for %#v at %s\n", err, args, c)
return false
}
}
return true
}