forked from TrueCloudLab/frostfs-node
011d0f605b
Continues replication executed in separate pool of goroutines, so there is no need in worker to handle replication tasks asynchronously. Signed-off-by: Alex Vanin <alexey@nspcc.ru>
131 lines
3 KiB
Go
131 lines
3 KiB
Go
package policer
|
|
|
|
import (
|
|
"context"
|
|
"errors"
|
|
"strings"
|
|
|
|
"github.com/nspcc-dev/neofs-node/pkg/core/container"
|
|
"github.com/nspcc-dev/neofs-node/pkg/local_object_storage/engine"
|
|
headsvc "github.com/nspcc-dev/neofs-node/pkg/services/object/head"
|
|
"github.com/nspcc-dev/neofs-node/pkg/services/replicator"
|
|
"github.com/nspcc-dev/neofs-sdk-go/netmap"
|
|
"github.com/nspcc-dev/neofs-sdk-go/object"
|
|
"go.uber.org/zap"
|
|
)
|
|
|
|
func (p *Policer) processObject(ctx context.Context, addr *object.Address) {
|
|
cnr, err := p.cnrSrc.Get(addr.ContainerID())
|
|
if err != nil {
|
|
p.log.Error("could not get container",
|
|
zap.Stringer("cid", addr.ContainerID()),
|
|
zap.String("error", err.Error()),
|
|
)
|
|
if errors.Is(err, container.ErrNotFound) {
|
|
prm := new(engine.InhumePrm)
|
|
prm.MarkAsGarbage(addr)
|
|
_, err := p.jobQueue.localStorage.Inhume(prm)
|
|
if err != nil {
|
|
p.log.Error("could not inhume object with missing container",
|
|
zap.Stringer("cid", addr.ContainerID()),
|
|
zap.Stringer("oid", addr.ObjectID()),
|
|
zap.String("error", err.Error()))
|
|
}
|
|
}
|
|
|
|
return
|
|
}
|
|
|
|
policy := cnr.PlacementPolicy()
|
|
|
|
nn, err := p.placementBuilder.BuildPlacement(addr, policy)
|
|
if err != nil {
|
|
p.log.Error("could not build placement vector for object",
|
|
zap.String("error", err.Error()),
|
|
)
|
|
|
|
return
|
|
}
|
|
|
|
replicas := policy.Replicas()
|
|
|
|
for i := range nn {
|
|
select {
|
|
case <-ctx.Done():
|
|
return
|
|
default:
|
|
}
|
|
|
|
p.processNodes(ctx, addr, nn[i], replicas[i].Count())
|
|
}
|
|
}
|
|
|
|
func (p *Policer) processNodes(ctx context.Context, addr *object.Address, nodes netmap.Nodes, shortage uint32) {
|
|
log := p.log.With(
|
|
zap.Stringer("object", addr),
|
|
)
|
|
|
|
prm := new(headsvc.RemoteHeadPrm).WithObjectAddress(addr)
|
|
redundantLocalCopy := false
|
|
|
|
for i := 0; i < len(nodes); i++ {
|
|
select {
|
|
case <-ctx.Done():
|
|
return
|
|
default:
|
|
}
|
|
|
|
if p.netmapKeys.IsLocalKey(nodes[i].PublicKey()) {
|
|
if shortage == 0 {
|
|
// we can call the redundant copy callback
|
|
// here to slightly improve the performance
|
|
// instead of readability.
|
|
redundantLocalCopy = true
|
|
break
|
|
} else {
|
|
shortage--
|
|
}
|
|
} else if shortage > 0 {
|
|
callCtx, cancel := context.WithTimeout(ctx, p.headTimeout)
|
|
|
|
_, err := p.remoteHeader.Head(callCtx, prm.WithNodeInfo(nodes[i].NodeInfo))
|
|
|
|
cancel()
|
|
|
|
if err != nil {
|
|
// FIXME: this is a temporary solution to resolve 404 response from remote node
|
|
// We need to distinguish problem nodes from nodes without an object.
|
|
if strings.Contains(err.Error(), headsvc.ErrNotFound.Error()) {
|
|
continue
|
|
} else {
|
|
log.Error("could not receive object header",
|
|
zap.String("error", err.Error()),
|
|
)
|
|
|
|
continue
|
|
}
|
|
} else {
|
|
shortage--
|
|
}
|
|
}
|
|
|
|
nodes = append(nodes[:i], nodes[i+1:]...)
|
|
i--
|
|
}
|
|
|
|
if shortage > 0 {
|
|
log.Info("shortage of object copies detected",
|
|
zap.Uint32("shortage", shortage),
|
|
)
|
|
|
|
p.replicator.HandleTask(ctx, new(replicator.Task).
|
|
WithObjectAddress(addr).
|
|
WithNodes(nodes).
|
|
WithCopiesNumber(shortage),
|
|
)
|
|
} else if redundantLocalCopy {
|
|
log.Info("redundant local object copy detected")
|
|
|
|
p.cbRedundantCopy(addr)
|
|
}
|
|
}
|