2018-11-20 19:30:15 +01:00
|
|
|
package job
|
|
|
|
|
|
|
|
import (
|
|
|
|
"context"
|
2020-04-11 15:49:41 +02:00
|
|
|
"fmt"
|
2019-03-17 21:06:19 +01:00
|
|
|
"sort"
|
|
|
|
|
2018-11-20 19:30:15 +01:00
|
|
|
"github.com/pkg/errors"
|
|
|
|
"github.com/prometheus/client_golang/prometheus"
|
2020-04-25 12:39:59 +02:00
|
|
|
"github.com/zrepl/zrepl/daemon/logging/trace"
|
2019-03-22 19:41:12 +01:00
|
|
|
|
2018-11-20 19:30:15 +01:00
|
|
|
"github.com/zrepl/zrepl/config"
|
|
|
|
"github.com/zrepl/zrepl/daemon/filters"
|
|
|
|
"github.com/zrepl/zrepl/daemon/job/wakeup"
|
|
|
|
"github.com/zrepl/zrepl/daemon/pruner"
|
|
|
|
"github.com/zrepl/zrepl/daemon/snapper"
|
|
|
|
"github.com/zrepl/zrepl/endpoint"
|
2019-03-17 21:06:19 +01:00
|
|
|
"github.com/zrepl/zrepl/replication/logic/pdu"
|
2018-11-20 19:30:15 +01:00
|
|
|
"github.com/zrepl/zrepl/zfs"
|
|
|
|
)
|
|
|
|
|
2018-11-21 01:54:56 +01:00
|
|
|
type SnapJob struct {
|
new features: {resumable,encrypted,hold-protected} send-recv, last-received-hold
- **Resumable Send & Recv Support**
No knobs required, automatically used where supported.
- **Hold-Protected Send & Recv**
Automatic ZFS holds to ensure that we can always resume a replication step.
- **Encrypted Send & Recv Support** for OpenZFS native encryption.
Configurable at the job level, i.e., for all filesystems a job is responsible for.
- **Receive-side hold on last received dataset**
The counterpart to the replication cursor bookmark on the send-side.
Ensures that incremental replication will always be possible between a sender and receiver.
Design Doc
----------
`replication/design.md` doc describes how we use ZFS holds and bookmarks to ensure that a single replication step is always resumable.
The replication algorithm described in the design doc introduces the notion of job IDs (please read the details on this design doc).
We reuse the job names for job IDs and use `JobID` type to ensure that a job name can be embedded into hold tags, bookmark names, etc.
This might BREAK CONFIG on upgrade.
Protocol Version Bump
---------------------
This commit makes backwards-incompatible changes to the replication/pdu protobufs.
Thus, bump the version number used in the protocol handshake.
Replication Cursor Format Change
--------------------------------
The new replication cursor bookmark format is: `#zrepl_CURSOR_G_${this.GUID}_J_${jobid}`
Including the GUID enables transaction-safe moving-forward of the cursor.
Including the job id enables that multiple sending jobs can send the same filesystem without interfering.
The `zrepl migrate replication-cursor:v1-v2` subcommand can be used to safely destroy old-format cursors once zrepl has created new-format cursors.
Changes in This Commit
----------------------
- package zfs
- infrastructure for holds
- infrastructure for resume token decoding
- implement a variant of OpenZFS's `entity_namecheck` and use it for validation in new code
- ZFSSendArgs to specify a ZFS send operation
- validation code protects against malicious resume tokens by checking that the token encodes the same send parameters that the send-side would use if no resume token were available (i.e. same filesystem, `fromguid`, `toguid`)
- RecvOptions support for `recv -s` flag
- convert a bunch of ZFS operations to be idempotent
- achieved through more differentiated error message scraping / additional pre-/post-checks
- package replication/pdu
- add field for encryption to send request messages
- add fields for resume handling to send & recv request messages
- receive requests now contain `FilesystemVersion To` in addition to the filesystem into which the stream should be `recv`d into
- can use `zfs recv $root_fs/$client_id/path/to/dataset@${To.Name}`, which enables additional validation after recv (i.e. whether `To.Guid` matched what we received in the stream)
- used to set `last-received-hold`
- package replication/logic
- introduce `PlannerPolicy` struct, currently only used to configure whether encrypted sends should be requested from the sender
- integrate encryption and resume token support into `Step` struct
- package endpoint
- move the concepts that endpoint builds on top of ZFS to a single file `endpoint/endpoint_zfs.go`
- step-holds + step-bookmarks
- last-received-hold
- new replication cursor + old replication cursor compat code
- adjust `endpoint/endpoint.go` handlers for
- encryption
- resumability
- new replication cursor
- last-received-hold
- client subcommand `zrepl holds list`: list all holds and hold-like bookmarks that zrepl thinks belong to it
- client subcommand `zrepl migrate replication-cursor:v1-v2`
2019-09-11 17:19:17 +02:00
|
|
|
name endpoint.JobID
|
2018-11-21 14:02:21 +01:00
|
|
|
fsfilter zfs.DatasetFilter
|
|
|
|
snapper *snapper.PeriodicOrManual
|
2018-11-20 19:30:15 +01:00
|
|
|
|
2019-03-17 21:18:25 +01:00
|
|
|
prunerFactory *pruner.LocalPrunerFactory
|
2018-11-20 19:30:15 +01:00
|
|
|
|
2019-03-17 21:07:42 +01:00
|
|
|
promPruneSecs *prometheus.HistogramVec // labels: prune_side
|
2018-11-20 19:30:15 +01:00
|
|
|
|
|
|
|
pruner *pruner.Pruner
|
|
|
|
}
|
|
|
|
|
new features: {resumable,encrypted,hold-protected} send-recv, last-received-hold
- **Resumable Send & Recv Support**
No knobs required, automatically used where supported.
- **Hold-Protected Send & Recv**
Automatic ZFS holds to ensure that we can always resume a replication step.
- **Encrypted Send & Recv Support** for OpenZFS native encryption.
Configurable at the job level, i.e., for all filesystems a job is responsible for.
- **Receive-side hold on last received dataset**
The counterpart to the replication cursor bookmark on the send-side.
Ensures that incremental replication will always be possible between a sender and receiver.
Design Doc
----------
`replication/design.md` doc describes how we use ZFS holds and bookmarks to ensure that a single replication step is always resumable.
The replication algorithm described in the design doc introduces the notion of job IDs (please read the details on this design doc).
We reuse the job names for job IDs and use `JobID` type to ensure that a job name can be embedded into hold tags, bookmark names, etc.
This might BREAK CONFIG on upgrade.
Protocol Version Bump
---------------------
This commit makes backwards-incompatible changes to the replication/pdu protobufs.
Thus, bump the version number used in the protocol handshake.
Replication Cursor Format Change
--------------------------------
The new replication cursor bookmark format is: `#zrepl_CURSOR_G_${this.GUID}_J_${jobid}`
Including the GUID enables transaction-safe moving-forward of the cursor.
Including the job id enables that multiple sending jobs can send the same filesystem without interfering.
The `zrepl migrate replication-cursor:v1-v2` subcommand can be used to safely destroy old-format cursors once zrepl has created new-format cursors.
Changes in This Commit
----------------------
- package zfs
- infrastructure for holds
- infrastructure for resume token decoding
- implement a variant of OpenZFS's `entity_namecheck` and use it for validation in new code
- ZFSSendArgs to specify a ZFS send operation
- validation code protects against malicious resume tokens by checking that the token encodes the same send parameters that the send-side would use if no resume token were available (i.e. same filesystem, `fromguid`, `toguid`)
- RecvOptions support for `recv -s` flag
- convert a bunch of ZFS operations to be idempotent
- achieved through more differentiated error message scraping / additional pre-/post-checks
- package replication/pdu
- add field for encryption to send request messages
- add fields for resume handling to send & recv request messages
- receive requests now contain `FilesystemVersion To` in addition to the filesystem into which the stream should be `recv`d into
- can use `zfs recv $root_fs/$client_id/path/to/dataset@${To.Name}`, which enables additional validation after recv (i.e. whether `To.Guid` matched what we received in the stream)
- used to set `last-received-hold`
- package replication/logic
- introduce `PlannerPolicy` struct, currently only used to configure whether encrypted sends should be requested from the sender
- integrate encryption and resume token support into `Step` struct
- package endpoint
- move the concepts that endpoint builds on top of ZFS to a single file `endpoint/endpoint_zfs.go`
- step-holds + step-bookmarks
- last-received-hold
- new replication cursor + old replication cursor compat code
- adjust `endpoint/endpoint.go` handlers for
- encryption
- resumability
- new replication cursor
- last-received-hold
- client subcommand `zrepl holds list`: list all holds and hold-like bookmarks that zrepl thinks belong to it
- client subcommand `zrepl migrate replication-cursor:v1-v2`
2019-09-11 17:19:17 +02:00
|
|
|
func (j *SnapJob) Name() string { return j.name.String() }
|
2018-11-20 19:30:15 +01:00
|
|
|
|
2018-11-21 01:54:56 +01:00
|
|
|
func (j *SnapJob) Type() Type { return TypeSnap }
|
2018-11-20 19:30:15 +01:00
|
|
|
|
2018-11-21 14:37:03 +01:00
|
|
|
func snapJobFromConfig(g *config.Global, in *config.SnapJob) (j *SnapJob, err error) {
|
2018-11-21 01:54:56 +01:00
|
|
|
j = &SnapJob{}
|
2018-11-21 14:02:21 +01:00
|
|
|
fsf, err := filters.DatasetMapFilterFromConfig(in.Filesystems)
|
|
|
|
if err != nil {
|
2019-12-11 18:00:00 +01:00
|
|
|
return nil, errors.Wrap(err, "cannot build filesystem filter")
|
2018-11-21 14:02:21 +01:00
|
|
|
}
|
|
|
|
j.fsfilter = fsf
|
|
|
|
|
|
|
|
if j.snapper, err = snapper.FromConfig(g, fsf, in.Snapshotting); err != nil {
|
|
|
|
return nil, errors.Wrap(err, "cannot build snapper")
|
|
|
|
}
|
new features: {resumable,encrypted,hold-protected} send-recv, last-received-hold
- **Resumable Send & Recv Support**
No knobs required, automatically used where supported.
- **Hold-Protected Send & Recv**
Automatic ZFS holds to ensure that we can always resume a replication step.
- **Encrypted Send & Recv Support** for OpenZFS native encryption.
Configurable at the job level, i.e., for all filesystems a job is responsible for.
- **Receive-side hold on last received dataset**
The counterpart to the replication cursor bookmark on the send-side.
Ensures that incremental replication will always be possible between a sender and receiver.
Design Doc
----------
`replication/design.md` doc describes how we use ZFS holds and bookmarks to ensure that a single replication step is always resumable.
The replication algorithm described in the design doc introduces the notion of job IDs (please read the details on this design doc).
We reuse the job names for job IDs and use `JobID` type to ensure that a job name can be embedded into hold tags, bookmark names, etc.
This might BREAK CONFIG on upgrade.
Protocol Version Bump
---------------------
This commit makes backwards-incompatible changes to the replication/pdu protobufs.
Thus, bump the version number used in the protocol handshake.
Replication Cursor Format Change
--------------------------------
The new replication cursor bookmark format is: `#zrepl_CURSOR_G_${this.GUID}_J_${jobid}`
Including the GUID enables transaction-safe moving-forward of the cursor.
Including the job id enables that multiple sending jobs can send the same filesystem without interfering.
The `zrepl migrate replication-cursor:v1-v2` subcommand can be used to safely destroy old-format cursors once zrepl has created new-format cursors.
Changes in This Commit
----------------------
- package zfs
- infrastructure for holds
- infrastructure for resume token decoding
- implement a variant of OpenZFS's `entity_namecheck` and use it for validation in new code
- ZFSSendArgs to specify a ZFS send operation
- validation code protects against malicious resume tokens by checking that the token encodes the same send parameters that the send-side would use if no resume token were available (i.e. same filesystem, `fromguid`, `toguid`)
- RecvOptions support for `recv -s` flag
- convert a bunch of ZFS operations to be idempotent
- achieved through more differentiated error message scraping / additional pre-/post-checks
- package replication/pdu
- add field for encryption to send request messages
- add fields for resume handling to send & recv request messages
- receive requests now contain `FilesystemVersion To` in addition to the filesystem into which the stream should be `recv`d into
- can use `zfs recv $root_fs/$client_id/path/to/dataset@${To.Name}`, which enables additional validation after recv (i.e. whether `To.Guid` matched what we received in the stream)
- used to set `last-received-hold`
- package replication/logic
- introduce `PlannerPolicy` struct, currently only used to configure whether encrypted sends should be requested from the sender
- integrate encryption and resume token support into `Step` struct
- package endpoint
- move the concepts that endpoint builds on top of ZFS to a single file `endpoint/endpoint_zfs.go`
- step-holds + step-bookmarks
- last-received-hold
- new replication cursor + old replication cursor compat code
- adjust `endpoint/endpoint.go` handlers for
- encryption
- resumability
- new replication cursor
- last-received-hold
- client subcommand `zrepl holds list`: list all holds and hold-like bookmarks that zrepl thinks belong to it
- client subcommand `zrepl migrate replication-cursor:v1-v2`
2019-09-11 17:19:17 +02:00
|
|
|
j.name, err = endpoint.MakeJobID(in.Name)
|
|
|
|
if err != nil {
|
|
|
|
return nil, errors.Wrap(err, "invalid job name")
|
|
|
|
}
|
2018-11-20 19:30:15 +01:00
|
|
|
j.promPruneSecs = prometheus.NewHistogramVec(prometheus.HistogramOpts{
|
|
|
|
Namespace: "zrepl",
|
|
|
|
Subsystem: "pruning",
|
|
|
|
Name: "time",
|
|
|
|
Help: "seconds spent in pruner",
|
new features: {resumable,encrypted,hold-protected} send-recv, last-received-hold
- **Resumable Send & Recv Support**
No knobs required, automatically used where supported.
- **Hold-Protected Send & Recv**
Automatic ZFS holds to ensure that we can always resume a replication step.
- **Encrypted Send & Recv Support** for OpenZFS native encryption.
Configurable at the job level, i.e., for all filesystems a job is responsible for.
- **Receive-side hold on last received dataset**
The counterpart to the replication cursor bookmark on the send-side.
Ensures that incremental replication will always be possible between a sender and receiver.
Design Doc
----------
`replication/design.md` doc describes how we use ZFS holds and bookmarks to ensure that a single replication step is always resumable.
The replication algorithm described in the design doc introduces the notion of job IDs (please read the details on this design doc).
We reuse the job names for job IDs and use `JobID` type to ensure that a job name can be embedded into hold tags, bookmark names, etc.
This might BREAK CONFIG on upgrade.
Protocol Version Bump
---------------------
This commit makes backwards-incompatible changes to the replication/pdu protobufs.
Thus, bump the version number used in the protocol handshake.
Replication Cursor Format Change
--------------------------------
The new replication cursor bookmark format is: `#zrepl_CURSOR_G_${this.GUID}_J_${jobid}`
Including the GUID enables transaction-safe moving-forward of the cursor.
Including the job id enables that multiple sending jobs can send the same filesystem without interfering.
The `zrepl migrate replication-cursor:v1-v2` subcommand can be used to safely destroy old-format cursors once zrepl has created new-format cursors.
Changes in This Commit
----------------------
- package zfs
- infrastructure for holds
- infrastructure for resume token decoding
- implement a variant of OpenZFS's `entity_namecheck` and use it for validation in new code
- ZFSSendArgs to specify a ZFS send operation
- validation code protects against malicious resume tokens by checking that the token encodes the same send parameters that the send-side would use if no resume token were available (i.e. same filesystem, `fromguid`, `toguid`)
- RecvOptions support for `recv -s` flag
- convert a bunch of ZFS operations to be idempotent
- achieved through more differentiated error message scraping / additional pre-/post-checks
- package replication/pdu
- add field for encryption to send request messages
- add fields for resume handling to send & recv request messages
- receive requests now contain `FilesystemVersion To` in addition to the filesystem into which the stream should be `recv`d into
- can use `zfs recv $root_fs/$client_id/path/to/dataset@${To.Name}`, which enables additional validation after recv (i.e. whether `To.Guid` matched what we received in the stream)
- used to set `last-received-hold`
- package replication/logic
- introduce `PlannerPolicy` struct, currently only used to configure whether encrypted sends should be requested from the sender
- integrate encryption and resume token support into `Step` struct
- package endpoint
- move the concepts that endpoint builds on top of ZFS to a single file `endpoint/endpoint_zfs.go`
- step-holds + step-bookmarks
- last-received-hold
- new replication cursor + old replication cursor compat code
- adjust `endpoint/endpoint.go` handlers for
- encryption
- resumability
- new replication cursor
- last-received-hold
- client subcommand `zrepl holds list`: list all holds and hold-like bookmarks that zrepl thinks belong to it
- client subcommand `zrepl migrate replication-cursor:v1-v2`
2019-09-11 17:19:17 +02:00
|
|
|
ConstLabels: prometheus.Labels{"zrepl_job": j.name.String()},
|
2018-11-21 04:26:03 +01:00
|
|
|
}, []string{"prune_side"})
|
2019-03-17 21:18:25 +01:00
|
|
|
j.prunerFactory, err = pruner.NewLocalPrunerFactory(in.Pruning, j.promPruneSecs)
|
2018-11-20 19:30:15 +01:00
|
|
|
if err != nil {
|
|
|
|
return nil, errors.Wrap(err, "cannot build snapjob pruning rules")
|
|
|
|
}
|
|
|
|
return j, nil
|
|
|
|
}
|
|
|
|
|
2018-11-21 01:54:56 +01:00
|
|
|
func (j *SnapJob) RegisterMetrics(registerer prometheus.Registerer) {
|
2018-11-20 19:30:15 +01:00
|
|
|
registerer.MustRegister(j.promPruneSecs)
|
|
|
|
}
|
|
|
|
|
2018-11-21 02:08:39 +01:00
|
|
|
type SnapJobStatus struct {
|
2019-09-19 00:13:55 +02:00
|
|
|
Pruning *pruner.Report
|
|
|
|
Snapshotting *snapper.Report // may be nil
|
2018-11-21 02:08:39 +01:00
|
|
|
}
|
|
|
|
|
2018-11-21 01:54:56 +01:00
|
|
|
func (j *SnapJob) Status() *Status {
|
2018-11-21 02:08:39 +01:00
|
|
|
s := &SnapJobStatus{}
|
2018-11-21 01:54:56 +01:00
|
|
|
t := j.Type()
|
2018-11-21 02:42:13 +01:00
|
|
|
if j.pruner != nil {
|
|
|
|
s.Pruning = j.pruner.Report()
|
2018-11-20 19:30:15 +01:00
|
|
|
}
|
2019-09-19 00:13:55 +02:00
|
|
|
s.Snapshotting = j.snapper.Report()
|
2018-11-20 19:30:15 +01:00
|
|
|
return &Status{Type: t, JobSpecific: s}
|
|
|
|
}
|
|
|
|
|
2019-03-20 23:01:24 +01:00
|
|
|
func (j *SnapJob) OwnedDatasetSubtreeRoot() (rfs *zfs.DatasetPath, ok bool) {
|
|
|
|
return nil, false
|
|
|
|
}
|
|
|
|
|
new features: {resumable,encrypted,hold-protected} send-recv, last-received-hold
- **Resumable Send & Recv Support**
No knobs required, automatically used where supported.
- **Hold-Protected Send & Recv**
Automatic ZFS holds to ensure that we can always resume a replication step.
- **Encrypted Send & Recv Support** for OpenZFS native encryption.
Configurable at the job level, i.e., for all filesystems a job is responsible for.
- **Receive-side hold on last received dataset**
The counterpart to the replication cursor bookmark on the send-side.
Ensures that incremental replication will always be possible between a sender and receiver.
Design Doc
----------
`replication/design.md` doc describes how we use ZFS holds and bookmarks to ensure that a single replication step is always resumable.
The replication algorithm described in the design doc introduces the notion of job IDs (please read the details on this design doc).
We reuse the job names for job IDs and use `JobID` type to ensure that a job name can be embedded into hold tags, bookmark names, etc.
This might BREAK CONFIG on upgrade.
Protocol Version Bump
---------------------
This commit makes backwards-incompatible changes to the replication/pdu protobufs.
Thus, bump the version number used in the protocol handshake.
Replication Cursor Format Change
--------------------------------
The new replication cursor bookmark format is: `#zrepl_CURSOR_G_${this.GUID}_J_${jobid}`
Including the GUID enables transaction-safe moving-forward of the cursor.
Including the job id enables that multiple sending jobs can send the same filesystem without interfering.
The `zrepl migrate replication-cursor:v1-v2` subcommand can be used to safely destroy old-format cursors once zrepl has created new-format cursors.
Changes in This Commit
----------------------
- package zfs
- infrastructure for holds
- infrastructure for resume token decoding
- implement a variant of OpenZFS's `entity_namecheck` and use it for validation in new code
- ZFSSendArgs to specify a ZFS send operation
- validation code protects against malicious resume tokens by checking that the token encodes the same send parameters that the send-side would use if no resume token were available (i.e. same filesystem, `fromguid`, `toguid`)
- RecvOptions support for `recv -s` flag
- convert a bunch of ZFS operations to be idempotent
- achieved through more differentiated error message scraping / additional pre-/post-checks
- package replication/pdu
- add field for encryption to send request messages
- add fields for resume handling to send & recv request messages
- receive requests now contain `FilesystemVersion To` in addition to the filesystem into which the stream should be `recv`d into
- can use `zfs recv $root_fs/$client_id/path/to/dataset@${To.Name}`, which enables additional validation after recv (i.e. whether `To.Guid` matched what we received in the stream)
- used to set `last-received-hold`
- package replication/logic
- introduce `PlannerPolicy` struct, currently only used to configure whether encrypted sends should be requested from the sender
- integrate encryption and resume token support into `Step` struct
- package endpoint
- move the concepts that endpoint builds on top of ZFS to a single file `endpoint/endpoint_zfs.go`
- step-holds + step-bookmarks
- last-received-hold
- new replication cursor + old replication cursor compat code
- adjust `endpoint/endpoint.go` handlers for
- encryption
- resumability
- new replication cursor
- last-received-hold
- client subcommand `zrepl holds list`: list all holds and hold-like bookmarks that zrepl thinks belong to it
- client subcommand `zrepl migrate replication-cursor:v1-v2`
2019-09-11 17:19:17 +02:00
|
|
|
func (j *SnapJob) SenderConfig() *endpoint.SenderConfig { return nil }
|
|
|
|
|
2018-11-21 01:54:56 +01:00
|
|
|
func (j *SnapJob) Run(ctx context.Context) {
|
2020-04-25 12:39:59 +02:00
|
|
|
ctx, endTask := trace.WithTaskAndSpan(ctx, "snap-job", j.Name())
|
2020-04-11 15:49:41 +02:00
|
|
|
defer endTask()
|
2018-11-20 19:30:15 +01:00
|
|
|
log := GetLogger(ctx)
|
|
|
|
|
|
|
|
defer log.Info("job exiting")
|
|
|
|
|
|
|
|
periodicDone := make(chan struct{})
|
|
|
|
ctx, cancel := context.WithCancel(ctx)
|
|
|
|
defer cancel()
|
2020-04-25 12:39:59 +02:00
|
|
|
periodicCtx, endTask := trace.WithTask(ctx, "snapshotting")
|
2020-04-11 15:49:41 +02:00
|
|
|
defer endTask()
|
|
|
|
go j.snapper.Run(periodicCtx, periodicDone)
|
2018-11-20 19:30:15 +01:00
|
|
|
|
|
|
|
invocationCount := 0
|
|
|
|
outer:
|
|
|
|
for {
|
|
|
|
log.Info("wait for wakeups")
|
|
|
|
select {
|
|
|
|
case <-ctx.Done():
|
|
|
|
log.WithError(ctx.Err()).Info("context")
|
|
|
|
break outer
|
|
|
|
|
|
|
|
case <-wakeup.Wait(ctx):
|
|
|
|
case <-periodicDone:
|
|
|
|
}
|
|
|
|
invocationCount++
|
2020-04-11 15:49:41 +02:00
|
|
|
|
2020-04-25 12:39:59 +02:00
|
|
|
invocationCtx, endSpan := trace.WithSpan(ctx, fmt.Sprintf("invocation-%d", invocationCount))
|
2020-04-11 15:49:41 +02:00
|
|
|
j.doPrune(invocationCtx)
|
|
|
|
endSpan()
|
2018-11-20 19:30:15 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-03-17 21:06:19 +01:00
|
|
|
// Adaptor that implements pruner.History around a pruner.Target.
|
|
|
|
// The ReplicationCursor method is Get-op only and always returns
|
|
|
|
// the filesystem's most recent version's GUID.
|
|
|
|
//
|
|
|
|
// TODO:
|
|
|
|
// This is a work-around for the current package daemon/pruner
|
|
|
|
// and package pruning.Snapshot limitation: they require the
|
|
|
|
// `Replicated` getter method be present, but obviously,
|
|
|
|
// a local job like SnapJob can't deliver on that.
|
|
|
|
// But the pruner.Pruner gives up on an FS if no replication
|
|
|
|
// cursor is present, which is why this pruner returns the
|
|
|
|
// most recent filesystem version.
|
|
|
|
type alwaysUpToDateReplicationCursorHistory struct {
|
2019-03-17 21:18:25 +01:00
|
|
|
// the Target passed as Target to BuildLocalPruner
|
2019-03-17 21:06:19 +01:00
|
|
|
target pruner.Target
|
|
|
|
}
|
|
|
|
|
|
|
|
var _ pruner.History = (*alwaysUpToDateReplicationCursorHistory)(nil)
|
|
|
|
|
|
|
|
func (h alwaysUpToDateReplicationCursorHistory) ReplicationCursor(ctx context.Context, req *pdu.ReplicationCursorReq) (*pdu.ReplicationCursorRes, error) {
|
|
|
|
fsvReq := &pdu.ListFilesystemVersionsReq{
|
|
|
|
Filesystem: req.GetFilesystem(),
|
|
|
|
}
|
|
|
|
res, err := h.target.ListFilesystemVersions(ctx, fsvReq)
|
|
|
|
if err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
|
|
|
fsvs := res.GetVersions()
|
|
|
|
if len(fsvs) <= 0 {
|
|
|
|
return &pdu.ReplicationCursorRes{Result: &pdu.ReplicationCursorRes_Notexist{Notexist: true}}, nil
|
|
|
|
}
|
|
|
|
// always return must recent version
|
|
|
|
sort.Slice(fsvs, func(i, j int) bool {
|
|
|
|
return fsvs[i].CreateTXG < fsvs[j].CreateTXG
|
|
|
|
})
|
|
|
|
mostRecent := fsvs[len(fsvs)-1]
|
|
|
|
return &pdu.ReplicationCursorRes{Result: &pdu.ReplicationCursorRes_Guid{Guid: mostRecent.GetGuid()}}, nil
|
|
|
|
}
|
|
|
|
|
|
|
|
func (h alwaysUpToDateReplicationCursorHistory) ListFilesystems(ctx context.Context, req *pdu.ListFilesystemReq) (*pdu.ListFilesystemRes, error) {
|
|
|
|
return h.target.ListFilesystems(ctx, req)
|
|
|
|
}
|
|
|
|
|
2018-11-21 02:42:13 +01:00
|
|
|
func (j *SnapJob) doPrune(ctx context.Context) {
|
2020-04-25 12:39:59 +02:00
|
|
|
ctx, endSpan := trace.WithSpan(ctx, "snap-job-do-prune")
|
2020-04-11 15:49:41 +02:00
|
|
|
defer endSpan()
|
2018-11-20 19:30:15 +01:00
|
|
|
log := GetLogger(ctx)
|
new features: {resumable,encrypted,hold-protected} send-recv, last-received-hold
- **Resumable Send & Recv Support**
No knobs required, automatically used where supported.
- **Hold-Protected Send & Recv**
Automatic ZFS holds to ensure that we can always resume a replication step.
- **Encrypted Send & Recv Support** for OpenZFS native encryption.
Configurable at the job level, i.e., for all filesystems a job is responsible for.
- **Receive-side hold on last received dataset**
The counterpart to the replication cursor bookmark on the send-side.
Ensures that incremental replication will always be possible between a sender and receiver.
Design Doc
----------
`replication/design.md` doc describes how we use ZFS holds and bookmarks to ensure that a single replication step is always resumable.
The replication algorithm described in the design doc introduces the notion of job IDs (please read the details on this design doc).
We reuse the job names for job IDs and use `JobID` type to ensure that a job name can be embedded into hold tags, bookmark names, etc.
This might BREAK CONFIG on upgrade.
Protocol Version Bump
---------------------
This commit makes backwards-incompatible changes to the replication/pdu protobufs.
Thus, bump the version number used in the protocol handshake.
Replication Cursor Format Change
--------------------------------
The new replication cursor bookmark format is: `#zrepl_CURSOR_G_${this.GUID}_J_${jobid}`
Including the GUID enables transaction-safe moving-forward of the cursor.
Including the job id enables that multiple sending jobs can send the same filesystem without interfering.
The `zrepl migrate replication-cursor:v1-v2` subcommand can be used to safely destroy old-format cursors once zrepl has created new-format cursors.
Changes in This Commit
----------------------
- package zfs
- infrastructure for holds
- infrastructure for resume token decoding
- implement a variant of OpenZFS's `entity_namecheck` and use it for validation in new code
- ZFSSendArgs to specify a ZFS send operation
- validation code protects against malicious resume tokens by checking that the token encodes the same send parameters that the send-side would use if no resume token were available (i.e. same filesystem, `fromguid`, `toguid`)
- RecvOptions support for `recv -s` flag
- convert a bunch of ZFS operations to be idempotent
- achieved through more differentiated error message scraping / additional pre-/post-checks
- package replication/pdu
- add field for encryption to send request messages
- add fields for resume handling to send & recv request messages
- receive requests now contain `FilesystemVersion To` in addition to the filesystem into which the stream should be `recv`d into
- can use `zfs recv $root_fs/$client_id/path/to/dataset@${To.Name}`, which enables additional validation after recv (i.e. whether `To.Guid` matched what we received in the stream)
- used to set `last-received-hold`
- package replication/logic
- introduce `PlannerPolicy` struct, currently only used to configure whether encrypted sends should be requested from the sender
- integrate encryption and resume token support into `Step` struct
- package endpoint
- move the concepts that endpoint builds on top of ZFS to a single file `endpoint/endpoint_zfs.go`
- step-holds + step-bookmarks
- last-received-hold
- new replication cursor + old replication cursor compat code
- adjust `endpoint/endpoint.go` handlers for
- encryption
- resumability
- new replication cursor
- last-received-hold
- client subcommand `zrepl holds list`: list all holds and hold-like bookmarks that zrepl thinks belong to it
- client subcommand `zrepl migrate replication-cursor:v1-v2`
2019-09-11 17:19:17 +02:00
|
|
|
sender := endpoint.NewSender(endpoint.SenderConfig{
|
|
|
|
JobID: j.name,
|
|
|
|
FSF: j.fsfilter,
|
|
|
|
// FIXME encryption setting is irrelevant for SnapJob because the endpoint is only used as pruner.Target
|
|
|
|
Encrypt: &zfs.NilBool{B: true},
|
|
|
|
})
|
2019-03-17 21:18:25 +01:00
|
|
|
j.pruner = j.prunerFactory.BuildLocalPruner(ctx, sender, alwaysUpToDateReplicationCursorHistory{sender})
|
2018-11-21 02:42:13 +01:00
|
|
|
log.Info("start pruning")
|
|
|
|
j.pruner.Prune()
|
|
|
|
log.Info("finished pruning")
|
2018-11-20 19:30:15 +01:00
|
|
|
}
|