2
2
mirror of https://github.com/octoleo/restic.git synced 2024-06-26 20:33:30 +00:00
restic/internal/restic/lock.go
Michael Eischer d92957dd78 lock: Implement strict lock expiry monitoring
Restic continued e.g. a backup task even when it failed to renew the
lock or failed to do so in time. For example if a backup client enters
standby during the backup this can allow other operations like `prune`
to run in the meantime (after calling `unlock`). After leaving standby
the backup client will continue its backup and upload indexes which
refer pack files that were removed in the meantime.

This commit introduces a goroutine explicitly monitoring for locks that
are not refreshed in time. To simplify the implementation there's now a
separate goroutine to refresh the lock and monitor for timeouts for each
lock. The monitoring goroutine would now cause the backup to fail as the
client has lost it's lock in the meantime.

The lock refresh goroutines are bound to the context used to lock the
repository initially. The context returned by `lockRepo` is also
cancelled when any of the goroutines exits. This ensures that the
context is cancelled whenever for any reason the lock is no longer
refreshed.
2022-10-03 00:19:46 +02:00

356 lines
9.0 KiB
Go

package restic
import (
"context"
"fmt"
"os"
"os/signal"
"os/user"
"sync"
"syscall"
"testing"
"time"
"github.com/restic/restic/internal/errors"
"golang.org/x/sync/errgroup"
"github.com/restic/restic/internal/debug"
)
// Lock represents a process locking the repository for an operation.
//
// There are two types of locks: exclusive and non-exclusive. There may be many
// different non-exclusive locks, but at most one exclusive lock, which can
// only be acquired while no non-exclusive lock is held.
//
// A lock must be refreshed regularly to not be considered stale, this must be
// triggered by regularly calling Refresh.
type Lock struct {
Time time.Time `json:"time"`
Exclusive bool `json:"exclusive"`
Hostname string `json:"hostname"`
Username string `json:"username"`
PID int `json:"pid"`
UID uint32 `json:"uid,omitempty"`
GID uint32 `json:"gid,omitempty"`
repo Repository
lockID *ID
}
// alreadyLockedError is returned when NewLock or NewExclusiveLock are unable to
// acquire the desired lock.
type alreadyLockedError struct {
otherLock *Lock
}
func (e *alreadyLockedError) Error() string {
s := ""
if e.otherLock.Exclusive {
s = "exclusively "
}
return fmt.Sprintf("repository is already locked %sby %v", s, e.otherLock)
}
// IsAlreadyLocked returns true iff err indicates that a repository is
// already locked.
func IsAlreadyLocked(err error) bool {
var e *alreadyLockedError
return errors.As(err, &e)
}
// NewLock returns a new, non-exclusive lock for the repository. If an
// exclusive lock is already held by another process, it returns an error
// that satisfies IsAlreadyLocked.
func NewLock(ctx context.Context, repo Repository) (*Lock, error) {
return newLock(ctx, repo, false)
}
// NewExclusiveLock returns a new, exclusive lock for the repository. If
// another lock (normal and exclusive) is already held by another process,
// it returns an error that satisfies IsAlreadyLocked.
func NewExclusiveLock(ctx context.Context, repo Repository) (*Lock, error) {
return newLock(ctx, repo, true)
}
var waitBeforeLockCheck = 200 * time.Millisecond
// TestSetLockTimeout can be used to reduce the lock wait timeout for tests.
func TestSetLockTimeout(t testing.TB, d time.Duration) {
t.Logf("setting lock timeout to %v", d)
waitBeforeLockCheck = d
}
func newLock(ctx context.Context, repo Repository, excl bool) (*Lock, error) {
lock := &Lock{
Time: time.Now(),
PID: os.Getpid(),
Exclusive: excl,
repo: repo,
}
hn, err := os.Hostname()
if err == nil {
lock.Hostname = hn
}
if err = lock.fillUserInfo(); err != nil {
return nil, err
}
if err = lock.checkForOtherLocks(ctx); err != nil {
return nil, err
}
lockID, err := lock.createLock(ctx)
if err != nil {
return nil, err
}
lock.lockID = &lockID
time.Sleep(waitBeforeLockCheck)
if err = lock.checkForOtherLocks(ctx); err != nil {
_ = lock.Unlock()
return nil, err
}
return lock, nil
}
func (l *Lock) fillUserInfo() error {
usr, err := user.Current()
if err != nil {
return nil
}
l.Username = usr.Username
l.UID, l.GID, err = uidGidInt(*usr)
return err
}
// checkForOtherLocks looks for other locks that currently exist in the repository.
//
// If an exclusive lock is to be created, checkForOtherLocks returns an error
// if there are any other locks, regardless if exclusive or not. If a
// non-exclusive lock is to be created, an error is only returned when an
// exclusive lock is found.
func (l *Lock) checkForOtherLocks(ctx context.Context) error {
return ForAllLocks(ctx, l.repo, l.lockID, func(id ID, lock *Lock, err error) error {
if err != nil {
// ignore locks that cannot be loaded
debug.Log("ignore lock %v: %v", id, err)
return nil
}
if l.Exclusive {
return &alreadyLockedError{otherLock: lock}
}
if !l.Exclusive && lock.Exclusive {
return &alreadyLockedError{otherLock: lock}
}
return nil
})
}
// createLock acquires the lock by creating a file in the repository.
func (l *Lock) createLock(ctx context.Context) (ID, error) {
id, err := SaveJSONUnpacked(ctx, l.repo, LockFile, l)
if err != nil {
return ID{}, err
}
return id, nil
}
// Unlock removes the lock from the repository.
func (l *Lock) Unlock() error {
if l == nil || l.lockID == nil {
return nil
}
return l.repo.Backend().Remove(context.TODO(), Handle{Type: LockFile, Name: l.lockID.String()})
}
var StaleLockTimeout = 30 * time.Minute
// Stale returns true if the lock is stale. A lock is stale if the timestamp is
// older than 30 minutes or if it was created on the current machine and the
// process isn't alive any more.
func (l *Lock) Stale() bool {
debug.Log("testing if lock %v for process %d is stale", l, l.PID)
if time.Since(l.Time) > StaleLockTimeout {
debug.Log("lock is stale, timestamp is too old: %v\n", l.Time)
return true
}
hn, err := os.Hostname()
if err != nil {
debug.Log("unable to find current hostname: %v", err)
// since we cannot find the current hostname, assume that the lock is
// not stale.
return false
}
if hn != l.Hostname {
// lock was created on a different host, assume the lock is not stale.
return false
}
// check if we can reach the process retaining the lock
exists := l.processExists()
if !exists {
debug.Log("could not reach process, %d, lock is probably stale\n", l.PID)
return true
}
debug.Log("lock not stale\n")
return false
}
// Refresh refreshes the lock by creating a new file in the backend with a new
// timestamp. Afterwards the old lock is removed.
func (l *Lock) Refresh(ctx context.Context) error {
debug.Log("refreshing lock %v", l.lockID)
l.Time = time.Now()
id, err := l.createLock(ctx)
if err != nil {
return err
}
debug.Log("new lock ID %v", id)
oldLockID := l.lockID
l.lockID = &id
return l.repo.Backend().Remove(context.TODO(), Handle{Type: LockFile, Name: oldLockID.String()})
}
func (l Lock) String() string {
text := fmt.Sprintf("PID %d on %s by %s (UID %d, GID %d)\nlock was created at %s (%s ago)\nstorage ID %v",
l.PID, l.Hostname, l.Username, l.UID, l.GID,
l.Time.Format("2006-01-02 15:04:05"), time.Since(l.Time),
l.lockID.Str())
return text
}
// listen for incoming SIGHUP and ignore
var ignoreSIGHUP sync.Once
func init() {
ignoreSIGHUP.Do(func() {
go func() {
c := make(chan os.Signal, 1)
signal.Notify(c, syscall.SIGHUP)
for s := range c {
debug.Log("Signal received: %v\n", s)
}
}()
})
}
// LoadLock loads and unserializes a lock from a repository.
func LoadLock(ctx context.Context, repo Repository, id ID) (*Lock, error) {
lock := &Lock{}
if err := LoadJSONUnpacked(ctx, repo, LockFile, id, lock); err != nil {
return nil, err
}
lock.lockID = &id
return lock, nil
}
// RemoveStaleLocks deletes all locks detected as stale from the repository.
func RemoveStaleLocks(ctx context.Context, repo Repository) (uint, error) {
var processed uint
err := ForAllLocks(ctx, repo, nil, func(id ID, lock *Lock, err error) error {
if err != nil {
// ignore locks that cannot be loaded
debug.Log("ignore lock %v: %v", id, err)
return nil
}
if lock.Stale() {
err = repo.Backend().Remove(ctx, Handle{Type: LockFile, Name: id.String()})
if err == nil {
processed++
}
return err
}
return nil
})
return processed, err
}
// RemoveAllLocks removes all locks forcefully.
func RemoveAllLocks(ctx context.Context, repo Repository) (uint, error) {
var processed uint
err := repo.List(ctx, LockFile, func(id ID, size int64) error {
err := repo.Backend().Remove(ctx, Handle{Type: LockFile, Name: id.String()})
if err == nil {
processed++
}
return err
})
return processed, err
}
// ForAllLocks reads all locks in parallel and calls the given callback.
// It is guaranteed that the function is not run concurrently. If the
// callback returns an error, this function is cancelled and also returns that error.
// If a lock ID is passed via excludeID, it will be ignored.
func ForAllLocks(ctx context.Context, repo Repository, excludeID *ID, fn func(ID, *Lock, error) error) error {
var m sync.Mutex
// track spawned goroutines using wg, create a new context which is
// cancelled as soon as an error occurs.
wg, ctx := errgroup.WithContext(ctx)
ch := make(chan ID)
// send list of lock files through ch, which is closed afterwards
wg.Go(func() error {
defer close(ch)
return repo.List(ctx, LockFile, func(id ID, size int64) error {
if excludeID != nil && id.Equal(*excludeID) {
return nil
}
select {
case <-ctx.Done():
return nil
case ch <- id:
}
return nil
})
})
// a worker receives an snapshot ID from ch, loads the snapshot
// and runs fn with id, the snapshot and the error
worker := func() error {
for id := range ch {
debug.Log("load lock %v", id)
lock, err := LoadLock(ctx, repo, id)
m.Lock()
err = fn(id, lock, err)
m.Unlock()
if err != nil {
return err
}
}
return nil
}
// For locks decoding is nearly for free, thus just assume were only limited by IO
for i := 0; i < int(repo.Connections()); i++ {
wg.Go(worker)
}
return wg.Wait()
}