zrepl/daemon/snapper/snapper.go

500 lines
12 KiB
Go
Raw Normal View History

2018-09-04 23:46:02 +02:00
package snapper
import (
"context"
"fmt"
"sort"
"sync"
2019-03-22 19:41:12 +01:00
"time"
2018-09-04 23:46:02 +02:00
2019-03-22 19:41:12 +01:00
"github.com/pkg/errors"
"github.com/zrepl/zrepl/daemon/logging/trace"
2019-03-22 19:41:12 +01:00
"github.com/zrepl/zrepl/config"
"github.com/zrepl/zrepl/daemon/hooks"
"github.com/zrepl/zrepl/daemon/logging"
2019-03-22 19:41:12 +01:00
"github.com/zrepl/zrepl/logger"
"github.com/zrepl/zrepl/util/envconst"
2019-03-22 19:41:12 +01:00
"github.com/zrepl/zrepl/zfs"
)
2018-09-04 23:46:02 +02:00
//go:generate stringer -type=SnapState
type SnapState uint
const (
SnapPending SnapState = 1 << iota
SnapStarted
SnapDone
SnapError
)
// All fields protected by Snapper.mtx
2018-09-04 23:46:02 +02:00
type snapProgress struct {
state SnapState
// SnapStarted, SnapDone, SnapError
name string
startAt time.Time
hookPlan *hooks.Plan
2018-09-04 23:46:02 +02:00
// SnapDone
doneAt time.Time
// SnapErr TODO disambiguate state
runResults hooks.PlanReport
2018-09-04 23:46:02 +02:00
}
type args struct {
ctx context.Context
prefix string
interval time.Duration
fsf zfs.DatasetFilter
2019-03-22 19:41:12 +01:00
snapshotsTaken chan<- struct{}
hooks *hooks.List
dryRun bool
2018-09-04 23:46:02 +02:00
}
type Snapper struct {
args args
2019-03-22 19:41:12 +01:00
mtx sync.Mutex
2018-09-04 23:46:02 +02:00
state State
// set in state Plan, used in Waiting
lastInvocation time.Time
// valid for state Snapshotting
plan map[*zfs.DatasetPath]*snapProgress
2018-09-04 23:46:02 +02:00
// valid for state SyncUp and Waiting
sleepUntil time.Time
// valid for state Err
err error
}
//go:generate stringer -type=State
type State uint
const (
2019-03-22 19:41:12 +01:00
SyncUp State = 1 << iota
SyncUpErrWait
2018-09-04 23:46:02 +02:00
Planning
Snapshotting
Waiting
ErrorWait
Stopped
2018-09-04 23:46:02 +02:00
)
func (s State) sf() state {
m := map[State]state{
2019-03-22 19:41:12 +01:00
SyncUp: syncUp,
SyncUpErrWait: wait,
2019-03-22 19:41:12 +01:00
Planning: plan,
Snapshotting: snapshot,
Waiting: wait,
ErrorWait: wait,
Stopped: nil,
2018-09-04 23:46:02 +02:00
}
return m[s]
}
type updater func(u func(*Snapper)) State
type state func(a args, u updater) state
type Logger = logger.Logger
func getLogger(ctx context.Context) Logger {
return logging.GetLogger(ctx, logging.SubsysSnapshot)
2018-09-04 23:46:02 +02:00
}
func PeriodicFromConfig(g *config.Global, fsf zfs.DatasetFilter, in *config.SnapshottingPeriodic) (*Snapper, error) {
if in.Prefix == "" {
2018-09-04 23:46:02 +02:00
return nil, errors.New("prefix must not be empty")
}
if in.Interval <= 0 {
return nil, errors.New("interval must be positive")
}
hookList, err := hooks.ListFromConfig(&in.Hooks)
if err != nil {
return nil, errors.Wrap(err, "hook config error")
}
2018-09-04 23:46:02 +02:00
args := args{
2019-03-22 19:41:12 +01:00
prefix: in.Prefix,
2018-09-04 23:46:02 +02:00
interval: in.Interval,
2019-03-22 19:41:12 +01:00
fsf: fsf,
hooks: hookList,
2018-09-04 23:46:02 +02:00
// ctx and log is set in Run()
}
return &Snapper{state: SyncUp, args: args}, nil
}
func (s *Snapper) Run(ctx context.Context, snapshotsTaken chan<- struct{}) {
defer trace.WithSpanFromStackUpdateCtx(&ctx)()
2018-09-04 23:46:02 +02:00
getLogger(ctx).Debug("start")
defer getLogger(ctx).Debug("stop")
s.args.snapshotsTaken = snapshotsTaken
s.args.ctx = ctx
s.args.dryRun = false // for future expansion
2018-09-04 23:46:02 +02:00
u := func(u func(*Snapper)) State {
s.mtx.Lock()
defer s.mtx.Unlock()
if u != nil {
u(s)
}
return s.state
}
var st state = syncUp
for st != nil {
pre := u(nil)
st = st(s.args, u)
post := u(nil)
getLogger(ctx).
WithField("transition", fmt.Sprintf("%s=>%s", pre, post)).
Debug("state transition")
}
}
func onErr(err error, u updater) state {
return u(func(s *Snapper) {
s.err = err
preState := s.state
switch s.state {
case SyncUp:
s.state = SyncUpErrWait
case Planning:
fallthrough
case Snapshotting:
s.state = ErrorWait
}
getLogger(s.args.ctx).WithError(err).WithField("pre_state", preState).WithField("post_state", s.state).Error("snapshotting error")
2018-09-04 23:46:02 +02:00
}).sf()
}
func onMainCtxDone(ctx context.Context, u updater) state {
return u(func(s *Snapper) {
s.err = ctx.Err()
s.state = Stopped
}).sf()
}
2018-09-04 23:46:02 +02:00
func syncUp(a args, u updater) state {
u(func(snapper *Snapper) {
snapper.lastInvocation = time.Now()
})
fss, err := listFSes(a.ctx, a.fsf)
2018-09-04 23:46:02 +02:00
if err != nil {
return onErr(err, u)
}
syncPoint, err := findSyncPoint(a.ctx, fss, a.prefix, a.interval)
2018-09-04 23:46:02 +02:00
if err != nil {
return onErr(err, u)
}
2019-03-22 19:41:12 +01:00
u(func(s *Snapper) {
2018-09-04 23:46:02 +02:00
s.sleepUntil = syncPoint
})
t := time.NewTimer(time.Until(syncPoint))
2018-09-04 23:46:02 +02:00
defer t.Stop()
select {
case <-t.C:
return u(func(s *Snapper) {
s.state = Planning
}).sf()
case <-a.ctx.Done():
return onMainCtxDone(a.ctx, u)
2018-09-04 23:46:02 +02:00
}
}
func plan(a args, u updater) state {
u(func(snapper *Snapper) {
snapper.lastInvocation = time.Now()
})
fss, err := listFSes(a.ctx, a.fsf)
2018-09-04 23:46:02 +02:00
if err != nil {
return onErr(err, u)
}
plan := make(map[*zfs.DatasetPath]*snapProgress, len(fss))
2018-09-04 23:46:02 +02:00
for _, fs := range fss {
plan[fs] = &snapProgress{state: SnapPending}
2018-09-04 23:46:02 +02:00
}
return u(func(s *Snapper) {
s.state = Snapshotting
s.plan = plan
s.err = nil
2018-09-04 23:46:02 +02:00
}).sf()
}
func snapshot(a args, u updater) state {
var plan map[*zfs.DatasetPath]*snapProgress
2018-09-04 23:46:02 +02:00
u(func(snapper *Snapper) {
plan = snapper.plan
})
hookMatchCount := make(map[hooks.Hook]int, len(*a.hooks))
for _, h := range *a.hooks {
hookMatchCount[h] = 0
}
anyFsHadErr := false
2018-09-04 23:46:02 +02:00
// TODO channel programs -> allow a little jitter?
for fs, progress := range plan {
suffix := time.Now().In(time.UTC).Format("20060102_150405_000")
snapname := fmt.Sprintf("%s%s", a.prefix, suffix)
ctx := logging.WithInjectedField(a.ctx, "fs", fs.ToString())
ctx = logging.WithInjectedField(ctx, "snap", snapname)
2018-09-04 23:46:02 +02:00
hookEnvExtra := hooks.Env{
hooks.EnvFS: fs.ToString(),
hooks.EnvSnapshot: snapname,
2018-09-04 23:46:02 +02:00
}
jobCallback := hooks.NewCallbackHookForFilesystem("snapshot", fs, func(ctx context.Context) (err error) {
l := getLogger(ctx)
l.Debug("create snapshot")
err = zfs.ZFSSnapshot(ctx, fs, snapname, false) // TODO propagate context to ZFSSnapshot
if err != nil {
l.WithError(err).Error("cannot create snapshot")
}
return
})
fsHadErr := false
var planReport hooks.PlanReport
var plan *hooks.Plan
{
filteredHooks, err := a.hooks.CopyFilteredForFilesystem(fs)
if err != nil {
getLogger(ctx).WithError(err).Error("unexpected filter error")
fsHadErr = true
goto updateFSState
}
// account for running hooks
for _, h := range filteredHooks {
hookMatchCount[h] = hookMatchCount[h] + 1
}
var planErr error
plan, planErr = hooks.NewPlan(&filteredHooks, hooks.PhaseSnapshot, jobCallback, hookEnvExtra)
if planErr != nil {
fsHadErr = true
getLogger(ctx).WithError(planErr).Error("cannot create job hook plan")
goto updateFSState
}
}
u(func(snapper *Snapper) {
progress.name = snapname
progress.startAt = time.Now()
progress.hookPlan = plan
progress.state = SnapStarted
})
{
getLogger(ctx).WithField("report", plan.Report().String()).Debug("begin run job plan")
plan.Run(ctx, a.dryRun)
planReport = plan.Report()
fsHadErr = planReport.HadError() // not just fatal errors
if fsHadErr {
getLogger(ctx).WithField("report", planReport.String()).Error("end run job plan with error")
} else {
getLogger(ctx).WithField("report", planReport.String()).Info("end run job plan successful")
}
}
updateFSState:
anyFsHadErr = anyFsHadErr || fsHadErr
2018-09-04 23:46:02 +02:00
u(func(snapper *Snapper) {
progress.doneAt = time.Now()
2018-09-04 23:46:02 +02:00
progress.state = SnapDone
if fsHadErr {
2018-09-04 23:46:02 +02:00
progress.state = SnapError
}
progress.runResults = planReport
2018-09-04 23:46:02 +02:00
})
}
select {
case a.snapshotsTaken <- struct{}{}:
default:
if a.snapshotsTaken != nil {
getLogger(a.ctx).Warn("callback channel is full, discarding snapshot update event")
}
2018-09-04 23:46:02 +02:00
}
for h, mc := range hookMatchCount {
if mc == 0 {
hookIdx := -1
for idx, ah := range *a.hooks {
if ah == h {
hookIdx = idx
break
}
}
getLogger(a.ctx).WithField("hook", h.String()).WithField("hook_number", hookIdx+1).Warn("hook did not match any snapshotted filesystems")
}
}
2018-09-04 23:46:02 +02:00
return u(func(snapper *Snapper) {
if anyFsHadErr {
snapper.state = ErrorWait
snapper.err = errors.New("one or more snapshots could not be created, check logs for details")
2018-09-04 23:46:02 +02:00
} else {
snapper.state = Waiting
snapper.err = nil
2018-09-04 23:46:02 +02:00
}
}).sf()
}
func wait(a args, u updater) state {
var sleepUntil time.Time
u(func(snapper *Snapper) {
lastTick := snapper.lastInvocation
snapper.sleepUntil = lastTick.Add(a.interval)
sleepUntil = snapper.sleepUntil
log := getLogger(a.ctx).WithField("sleep_until", sleepUntil).WithField("duration", a.interval)
logFunc := log.Debug
if snapper.state == ErrorWait || snapper.state == SyncUpErrWait {
logFunc = log.Error
}
logFunc("enter wait-state after error")
2018-09-04 23:46:02 +02:00
})
t := time.NewTimer(time.Until(sleepUntil))
2018-09-04 23:46:02 +02:00
defer t.Stop()
select {
case <-t.C:
return u(func(snapper *Snapper) {
snapper.state = Planning
}).sf()
case <-a.ctx.Done():
return onMainCtxDone(a.ctx, u)
2018-09-04 23:46:02 +02:00
}
}
func listFSes(ctx context.Context, mf zfs.DatasetFilter) (fss []*zfs.DatasetPath, err error) {
return zfs.ZFSListMapping(ctx, mf)
2018-09-04 23:46:02 +02:00
}
var syncUpWarnNoSnapshotUntilSyncupMinDuration = envconst.Duration("ZREPL_SNAPPER_SYNCUP_WARN_MIN_DURATION", 1*time.Second)
// see docs/snapshotting.rst
func findSyncPoint(ctx context.Context, fss []*zfs.DatasetPath, prefix string, interval time.Duration) (syncPoint time.Time, err error) {
const (
prioHasVersions int = iota
prioNoVersions
)
2018-09-04 23:46:02 +02:00
type snapTime struct {
ds *zfs.DatasetPath
prio int // lower is higher
2018-09-04 23:46:02 +02:00
time time.Time
}
if len(fss) == 0 {
return time.Now(), nil
}
snaptimes := make([]snapTime, 0, len(fss))
hardErrs := 0
2018-09-04 23:46:02 +02:00
now := time.Now()
getLogger(ctx).Debug("examine filesystem state to find sync point")
2018-09-04 23:46:02 +02:00
for _, d := range fss {
ctx := logging.WithInjectedField(ctx, "fs", d.ToString())
syncPoint, err := findSyncPointFSNextOptimalSnapshotTime(ctx, now, interval, prefix, d)
if err == findSyncPointFSNoFilesystemVersionsErr {
snaptimes = append(snaptimes, snapTime{
ds: d,
prio: prioNoVersions,
time: now,
})
} else if err != nil {
hardErrs++
getLogger(ctx).WithError(err).Error("cannot determine optimal sync point for this filesystem")
} else {
getLogger(ctx).WithField("syncPoint", syncPoint).Debug("found optimal sync point for this filesystem")
snaptimes = append(snaptimes, snapTime{
ds: d,
prio: prioHasVersions,
time: syncPoint,
})
2018-09-04 23:46:02 +02:00
}
}
2018-09-04 23:46:02 +02:00
if hardErrs == len(fss) {
return time.Time{}, fmt.Errorf("hard errors in determining sync point for every matching filesystem")
}
2018-09-04 23:46:02 +02:00
if len(snaptimes) == 0 {
panic("implementation error: loop must either inc hardErrs or add result to snaptimes")
}
2018-09-04 23:46:02 +02:00
// sort ascending by (prio,time)
// => those filesystems with versions win over those without any
sort.Slice(snaptimes, func(i, j int) bool {
if snaptimes[i].prio == snaptimes[j].prio {
return snaptimes[i].time.Before(snaptimes[j].time)
2018-09-04 23:46:02 +02:00
}
return snaptimes[i].prio < snaptimes[j].prio
})
winnerSyncPoint := snaptimes[0].time
l := getLogger(ctx).WithField("syncPoint", winnerSyncPoint.String())
l.Info("determined sync point")
if winnerSyncPoint.Sub(now) > syncUpWarnNoSnapshotUntilSyncupMinDuration {
for _, st := range snaptimes {
if st.prio == prioNoVersions {
l.WithField("fs", st.ds.ToString()).Warn("filesystem will not be snapshotted until sync point")
}
2018-09-04 23:46:02 +02:00
}
}
return snaptimes[0].time, nil
}
var findSyncPointFSNoFilesystemVersionsErr = fmt.Errorf("no filesystem versions")
func findSyncPointFSNextOptimalSnapshotTime(ctx context.Context, now time.Time, interval time.Duration, prefix string, d *zfs.DatasetPath) (time.Time, error) {
fsvs, err := zfs.ZFSListFilesystemVersions(ctx, d, zfs.ListFilesystemVersionsOptions{
endpoint: refactor, fix stale holds on initial replication failure, zfs-abstractions subcmd, more efficient ZFS queries The motivation for this recatoring are based on two independent issues: - @JMoVS found that the changes merged as part of #259 slowed his OS X based installation down significantly. Analysis of the zfs command logging introduced in #296 showed that `zfs holds` took most of the execution time, and they pointed out that not all of those `zfs holds` invocations were actually necessary. I.e.: zrepl was inefficient about retrieving information from ZFS. - @InsanePrawn found that failures on initial replication would lead to step holds accumulating on the sending side, i.e. they would never be cleaned up in the HintMostRecentCommonAncestor RPC handler. That was because we only sent that RPC if there was a most recent common ancestor detected during replication planning. @InsanePrawn prototyped an implementation of a `zrepl zfs-abstractions release` command to mitigate the situation. As part of that development work and back-and-forth with @problame, it became evident that the abstractions that #259 built on top of zfs in package endpoint (step holds, replication cursor, last-received-hold), were not well-represented for re-use in the `zrepl zfs-abstractions release` subocommand prototype. This commit refactors package endpoint to address both of these issues: - endpoint abstractions now share an interface `Abstraction` that, among other things, provides a uniform `Destroy()` method. However, that method should not be destroyed directly but instead the package-level `BatchDestroy` function should be used in order to allow for a migration to zfs channel programs in the future. - endpoint now has a query facitilty (`ListAbstractions`) which is used to find on-disk - step holds and bookmarks - replication cursors (v1, v2) - last-received-holds By describing the query in a struct, we can centralized the retrieval of information via the ZFS CLI and only have to be clever once. We are "clever" in the following ways: - When asking for hold-based abstractions, we only run `zfs holds` on snapshot that have `userrefs` > 0 - To support this functionality, add field `UserRefs` to zfs.FilesystemVersion and retrieve it anywhere we retrieve zfs.FilesystemVersion from ZFS. - When asking only for bookmark-based abstractions, we only run `zfs list -t bookmark`, not with snapshots. - Currently unused (except for CLI) per-filesystem concurrent lookup - Option to only include abstractions with CreateTXG in a specified range - refactor `endpoint`'s various ZFS info retrieval methods to use `ListAbstractions` - rename the `zrepl holds list` command to `zrepl zfs-abstractions list` - make `zrepl zfs-abstractions list` consume endpoint.ListAbstractions - Add a `ListStale` method which, given a query template, lists stale holds and bookmarks. - it uses replication cursor has different modes - the new `zrepl zfs-abstractions release-{all,stale}` commands can be used to remove abstractions of package endpoint - Adjust HintMostRecentCommonAncestor RPC for stale-holds cleanup: - send it also if no most recent common ancestor exists between sender and receiver - have the sender clean up its abstractions when it receives the RPC with no most recent common ancestor, using `ListStale` - Due to changed semantics, bump the protocol version. - Adjust HintMostRecentCommonAncestor RPC for performance problems encountered by @JMoVS - by default, per (job,fs)-combination, only consider cleaning step holds in the createtxg range `[last replication cursor,conservatively-estimated-receive-side-version)` - this behavior ensures resumability at cost proportional to the time that replication was donw - however, as explained in a comment, we might leak holds if the zrepl daemon stops running - that trade-off is acceptable because in the presumably rare this might happen the user has two tools at their hand: - Tool 1: run `zrepl zfs-abstractions release-stale` - Tool 2: use env var `ZREPL_ENDPOINT_SENDER_HINT_MOST_RECENT_STEP_HOLD_CLEANUP_MODE` to adjust the lower bound of the createtxg range (search for it in the code). The env var can also be used to disable hold-cleanup on the send-side entirely. supersedes closes #293 supersedes closes #282 fixes #280 fixes #278 Additionaly, we fixed a couple of bugs: - zfs: fix half-nil error reporting of dataset-does-not-exist for ZFSListChan and ZFSBookmark - endpoint: Sender's `HintMostRecentCommonAncestor` handler would not check whether access to the specified filesystem was allowed.
2020-03-26 23:43:17 +01:00
Types: zfs.Snapshots,
ShortnamePrefix: prefix,
})
if err != nil {
return time.Time{}, errors.Wrap(err, "list filesystem versions")
}
if len(fsvs) <= 0 {
return time.Time{}, findSyncPointFSNoFilesystemVersionsErr
2018-09-04 23:46:02 +02:00
}
// Sort versions by creation
sort.SliceStable(fsvs, func(i, j int) bool {
return fsvs[i].CreateTXG < fsvs[j].CreateTXG
2018-09-04 23:46:02 +02:00
})
latest := fsvs[len(fsvs)-1]
getLogger(ctx).WithField("creation", latest.Creation).Debug("found latest snapshot")
since := now.Sub(latest.Creation)
if since < 0 {
return time.Time{}, fmt.Errorf("snapshot %q is from the future: creation=%q now=%q", latest.ToAbsPath(d), latest.Creation, now)
}
2018-09-04 23:46:02 +02:00
return latest.Creation.Add(interval), nil
2018-09-04 23:46:02 +02:00
}