ipfs-cluster/state/dsstate/datastore.go

321 lines
7.7 KiB
Go
Raw Normal View History

// Package dsstate implements the IPFS Cluster state interface using
// an underlying go-datastore.
package dsstate
import (
"context"
"io"
"github.com/ipfs/ipfs-cluster/api"
"github.com/ipfs/ipfs-cluster/state"
cid "github.com/ipfs/go-cid"
ds "github.com/ipfs/go-datastore"
query "github.com/ipfs/go-datastore/query"
dshelp "github.com/ipfs/go-ipfs-ds-help"
logging "github.com/ipfs/go-log/v2"
codec "github.com/ugorji/go/codec"
trace "go.opencensus.io/trace"
)
var _ state.State = (*State)(nil)
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
var _ state.BatchingState = (*BatchingState)(nil)
var logger = logging.Logger("dsstate")
// State implements the IPFS Cluster "state" interface by wrapping
// a go-datastore and choosing how api.Pin objects are stored
// in it. It also provides serialization methods for the whole
// state which are datastore-independent.
type State struct {
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
dsRead ds.Read
dsWrite ds.Write
codecHandle codec.Handle
namespace ds.Key
// version int
}
// DefaultHandle returns the codec handler of choice (Msgpack).
func DefaultHandle() codec.Handle {
h := &codec.MsgpackHandle{}
return h
}
// New returns a new state using the given datastore.
//
// All keys are namespaced with the given string when written. Thus the same
// go-datastore can be sharded for different uses.
//
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
// The Handle controls options for the serialization of the full state
// (marshaling/unmarshaling).
func New(dstore ds.Datastore, namespace string, handle codec.Handle) (*State, error) {
if handle == nil {
handle = DefaultHandle()
}
st := &State{
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
dsRead: dstore,
dsWrite: dstore,
codecHandle: handle,
namespace: ds.NewKey(namespace),
}
return st, nil
}
// Add adds a new Pin or replaces an existing one.
func (st *State) Add(ctx context.Context, c *api.Pin) error {
_, span := trace.StartSpan(ctx, "state/dsstate/Add")
defer span.End()
ps, err := st.serializePin(c)
if err != nil {
return err
}
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
return st.dsWrite.Put(st.key(c.Cid), ps)
}
// Rm removes an existing Pin. It is a no-op when the
// item does not exist.
func (st *State) Rm(ctx context.Context, c cid.Cid) error {
_, span := trace.StartSpan(ctx, "state/dsstate/Rm")
defer span.End()
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
err := st.dsWrite.Delete(st.key(c))
if err == ds.ErrNotFound {
return nil
}
return err
}
// Get returns a Pin from the store and whether it
// was present. When not present, a default pin
// is returned.
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
func (st *State) Get(ctx context.Context, c cid.Cid) (*api.Pin, error) {
_, span := trace.StartSpan(ctx, "state/dsstate/Get")
defer span.End()
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
v, err := st.dsRead.Get(st.key(c))
if err != nil {
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
if err == ds.ErrNotFound {
return nil, state.ErrNotFound
}
return nil, err
}
p, err := st.deserializePin(c, v)
if err != nil {
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
return nil, err
}
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
return p, nil
}
// Has returns whether a Cid is stored.
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
func (st *State) Has(ctx context.Context, c cid.Cid) (bool, error) {
_, span := trace.StartSpan(ctx, "state/dsstate/Has")
defer span.End()
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
ok, err := st.dsRead.Has(st.key(c))
if err != nil {
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
return false, err
}
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
return ok, nil
}
// List returns the unsorted list of all Pins that have been added to the
// datastore.
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
func (st *State) List(ctx context.Context) ([]*api.Pin, error) {
_, span := trace.StartSpan(ctx, "state/dsstate/List")
defer span.End()
q := query.Query{
Prefix: st.namespace.String(),
}
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
results, err := st.dsRead.Query(q)
if err != nil {
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
return nil, err
}
defer results.Close()
var pins []*api.Pin
for r := range results.Next() {
if r.Error != nil {
logger.Errorf("error in query result: %s", r.Error)
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
return pins, r.Error
}
k := ds.NewKey(r.Key)
ci, err := st.unkey(k)
if err != nil {
logger.Warn("bad key (ignoring). key: ", k, "error: ", err)
continue
}
p, err := st.deserializePin(ci, r.Value)
if err != nil {
logger.Errorf("error deserializing pin (%s): %s", r.Key, err)
continue
}
pins = append(pins, p)
}
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
return pins, nil
}
// Migrate migrates an older state version to the current one.
// This is a no-op for now.
func (st *State) Migrate(ctx context.Context, r io.Reader) error {
return nil
}
type serialEntry struct {
Key string `codec:"k"`
Value []byte `codec:"v"`
}
// Marshal dumps the state to a writer. It does this by encoding every
// key/value in the store. The keys are stored without the namespace part to
// reduce the size of the snapshot.
func (st *State) Marshal(w io.Writer) error {
q := query.Query{
Prefix: st.namespace.String(),
}
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
results, err := st.dsRead.Query(q)
if err != nil {
return err
}
defer results.Close()
enc := codec.NewEncoder(w, st.codecHandle)
for r := range results.Next() {
if r.Error != nil {
logger.Errorf("error in query result: %s", r.Error)
return r.Error
}
k := ds.NewKey(r.Key)
// reduce snapshot size by not storing the prefix
err := enc.Encode(serialEntry{
Key: k.BaseNamespace(),
Value: r.Value,
})
if err != nil {
logger.Error(err)
return err
}
}
return nil
}
// Unmarshal reads and parses a previous dump of the state.
// All the parsed key/values are added to the store. As of now,
// Unmarshal does not empty the existing store from any values
// before unmarshaling from the given reader.
func (st *State) Unmarshal(r io.Reader) error {
dec := codec.NewDecoder(r, st.codecHandle)
for {
var entry serialEntry
if err := dec.Decode(&entry); err == io.EOF {
break
} else if err != nil {
return err
}
k := st.namespace.Child(ds.NewKey(entry.Key))
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
err := st.dsWrite.Put(k, entry.Value)
if err != nil {
logger.Error("error adding unmarshaled key to datastore:", err)
return err
}
}
return nil
}
// used to be on go-ipfs-ds-help
func cidToDsKey(c cid.Cid) ds.Key {
return dshelp.NewKeyFromBinary(c.Bytes())
}
// used to be on go-ipfs-ds-help
func dsKeyToCid(k ds.Key) (cid.Cid, error) {
kb, err := dshelp.BinaryFromDsKey(k)
if err != nil {
return cid.Undef, err
}
return cid.Cast(kb)
}
// convert Cid to /namespace/cid1Key
func (st *State) key(c cid.Cid) ds.Key {
k := cidToDsKey(c)
return st.namespace.Child(k)
}
// convert /namespace/cidKey to Cid
func (st *State) unkey(k ds.Key) (cid.Cid, error) {
return dsKeyToCid(ds.NewKey(k.BaseNamespace()))
}
// this decides how a Pin object is serialized to be stored in the
// datastore. Changing this may require a migration!
func (st *State) serializePin(c *api.Pin) ([]byte, error) {
return c.ProtoMarshal()
}
// this deserializes a Pin object from the datastore. It should be
// the exact opposite from serializePin.
func (st *State) deserializePin(c cid.Cid, buf []byte) (*api.Pin, error) {
p := &api.Pin{}
err := p.ProtoUnmarshal(buf)
p.Cid = c
return p, err
}
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
// BatchingState implements the IPFS Cluster "state" interface by wrapping a
// batching go-datastore. All writes are batched and only written disk
// when Commit() is called.
type BatchingState struct {
*State
batch ds.Batch
}
// NewBatching returns a new batching statate using the given datastore.
//
// All keys are namespaced with the given string when written. Thus the same
// go-datastore can be sharded for different uses.
//
// The Handle controls options for the serialization of the full state
// (marshaling/unmarshaling).
func NewBatching(dstore ds.Batching, namespace string, handle codec.Handle) (*BatchingState, error) {
if handle == nil {
handle = DefaultHandle()
}
batch, err := dstore.Batch()
if err != nil {
return nil, err
}
st := &State{
dsRead: dstore,
dsWrite: batch,
codecHandle: handle,
namespace: ds.NewKey(namespace),
}
bst := &BatchingState{}
bst.State = st
bst.batch = batch
return bst, nil
}
// Commit persists the batched write operations.
func (bst *BatchingState) Commit(ctx context.Context) error {
_, span := trace.StartSpan(ctx, "state/dsstate/Commit")
Consensus: add new "crdt" consensus component This adds a new "crdt" consensus component using go-ds-crdt. This implies several refactors to fully make cluster consensus-component independent: * Delete mapstate and fully adopt dsstate (after people have migrated). * Return errors from state methods rather than ignoring them. * Add a new "datastore" modules so that we can configure datastores in the main configuration like other components. * Let the consensus components fully define the "state.State". Thus, they do not receive the state, they receive the storage where we put the state (a go-datastore). * Allow to customize how the monitor component obtains Peers() (the current peerset), including avoiding using the current peerset. At the moment the crdt consensus uses the monitoring component to define the current peerset. Therefore the monitor component cannot rely on the consensus component to produce a peerset. * Re-factor/re-implementation of "ipfs-cluster-service state" operations. Includes the dissapearance of the "migrate" one. The CRDT consensus component defines creates a crdt-datastore (with ipfs-lite) and uses it to intitialize a dssate. Thus the crdt-store is elegantly wrapped. Any modifications to the state get automatically replicated to other peers. We store all the CRDT DAG blocks in the local datastore. The consensus components only expose a ReadOnly state, as any modifications to the shared state should happen through them. DHT and PubSub facilities must now be created outside of Cluster and passed in so they can be re-used by different components.
2019-02-20 14:24:25 +00:00
defer span.End()
return bst.batch.Commit()
}