forked from TrueCloudLab/frostfs-node
[#1681] node: Block only Object service ops under maintenance
In previous implementation node blocked any operation of local object storage in maintenance mode. There is a need to perform some storage operations like data evacuation or restoration. Do not call block storage engine in maintenance mode. Make all Object service operations to return `apistatus.NodeUnderMaintenance` error from each local op. Signed-off-by: Leonard Lyubich <ctulhurider@gmail.com>
This commit is contained in:
parent
082602b668
commit
713aea06fa
8 changed files with 86 additions and 31 deletions
|
@ -335,6 +335,28 @@ type cfg struct {
|
|||
|
||||
// current network map
|
||||
netMap atomicstd.Value // type netmap.NetMap
|
||||
|
||||
// is node under maintenance
|
||||
isMaintenance atomic.Bool
|
||||
}
|
||||
|
||||
// starts node's maintenance.
|
||||
func (c *cfg) startMaintenance() {
|
||||
c.isMaintenance.Store(true)
|
||||
c.log.Info("started local node's maintenance")
|
||||
}
|
||||
|
||||
// stops node's maintenance.
|
||||
func (c *cfg) stopMaintenance() {
|
||||
c.isMaintenance.Store(false)
|
||||
c.log.Info("stopped local node's maintenance")
|
||||
}
|
||||
|
||||
// IsMaintenance checks if storage node is under maintenance.
|
||||
//
|
||||
// Provides util.NodeState to Object service.
|
||||
func (c *cfg) IsMaintenance() bool {
|
||||
return c.isMaintenance.Load()
|
||||
}
|
||||
|
||||
// ReadCurrentNetMap reads network map which has been cached at the
|
||||
|
|
|
@ -16,7 +16,6 @@ import (
|
|||
netmapTransportGRPC "github.com/nspcc-dev/neofs-node/pkg/network/transport/netmap/grpc"
|
||||
"github.com/nspcc-dev/neofs-node/pkg/services/control"
|
||||
netmapService "github.com/nspcc-dev/neofs-node/pkg/services/netmap"
|
||||
apistatus "github.com/nspcc-dev/neofs-sdk-go/client/status"
|
||||
netmapSDK "github.com/nspcc-dev/neofs-sdk-go/netmap"
|
||||
subnetid "github.com/nspcc-dev/neofs-sdk-go/subnet/id"
|
||||
"github.com/nspcc-dev/neofs-sdk-go/version"
|
||||
|
@ -319,30 +318,17 @@ func addNewEpochAsyncNotificationHandler(c *cfg, h event.Handler) {
|
|||
|
||||
var errRelayBootstrap = errors.New("setting netmap status is forbidden in relay mode")
|
||||
|
||||
var errNodeMaintenance apistatus.NodeUnderMaintenance
|
||||
|
||||
func (c *cfg) SetNetmapStatus(st control.NetmapStatus) error {
|
||||
switch st {
|
||||
default:
|
||||
return fmt.Errorf("unsupported status %v", st)
|
||||
case control.NetmapStatus_MAINTENANCE:
|
||||
err := c.cfgObject.cfgLocalStorage.localStorage.BlockExecution(errNodeMaintenance)
|
||||
if err != nil {
|
||||
return fmt.Errorf("block execution of local object storage: %w", err)
|
||||
}
|
||||
|
||||
// TODO: #1691 think how to process two different actions which can fail both
|
||||
|
||||
c.startMaintenance()
|
||||
return c.updateNetMapState((*nmClient.UpdatePeerPrm).SetMaintenance)
|
||||
case control.NetmapStatus_ONLINE, control.NetmapStatus_OFFLINE:
|
||||
}
|
||||
|
||||
err := c.cfgObject.cfgLocalStorage.localStorage.ResumeExecution()
|
||||
if err != nil {
|
||||
c.log.Error("failed to resume local object operations",
|
||||
zap.String("error", err.Error()),
|
||||
)
|
||||
}
|
||||
c.stopMaintenance()
|
||||
|
||||
if !c.needBootstrap() {
|
||||
return errRelayBootstrap
|
||||
|
|
|
@ -253,18 +253,18 @@ func initObjectService(c *cfg) {
|
|||
|
||||
c.workers = append(c.workers, pol)
|
||||
|
||||
var os putsvc.ObjectStorage
|
||||
var os putsvc.ObjectStorage = engineWithoutNotifications{
|
||||
e: ls,
|
||||
state: c,
|
||||
}
|
||||
|
||||
if c.cfgNotifications.enabled {
|
||||
os = engineWithNotifications{
|
||||
e: ls,
|
||||
base: os,
|
||||
nw: c.cfgNotifications.nw,
|
||||
ns: c.cfgNetmap.state,
|
||||
defaultTopic: c.cfgNotifications.defaultTopic,
|
||||
}
|
||||
} else {
|
||||
os = engineWithoutNotifications{
|
||||
e: ls,
|
||||
}
|
||||
}
|
||||
|
||||
sPut := putsvc.NewService(
|
||||
|
@ -291,7 +291,7 @@ func initObjectService(c *cfg) {
|
|||
|
||||
sSearch := searchsvc.New(
|
||||
searchsvc.WithLogger(c.log),
|
||||
searchsvc.WithLocalStorageEngine(ls),
|
||||
searchsvc.WithLocalStorageEngine(ls, c),
|
||||
searchsvc.WithClientConstructor(coreConstructor),
|
||||
searchsvc.WithTraverserGenerator(
|
||||
traverseGen.WithTraverseOptions(
|
||||
|
@ -318,6 +318,7 @@ func initObjectService(c *cfg) {
|
|||
),
|
||||
getsvc.WithNetMapSource(c.netMapSource),
|
||||
getsvc.WithKeyStorage(keyStorage),
|
||||
getsvc.WithNodeState(c),
|
||||
)
|
||||
|
||||
*c.cfgObject.getSvc = *sGet // need smth better
|
||||
|
@ -552,15 +553,15 @@ func (c *reputationClientConstructor) Get(info coreclient.NodeInfo) (coreclient.
|
|||
}
|
||||
|
||||
type engineWithNotifications struct {
|
||||
e *engine.StorageEngine
|
||||
nw notificationWriter
|
||||
ns netmap.State
|
||||
base putsvc.ObjectStorage
|
||||
nw notificationWriter
|
||||
ns netmap.State
|
||||
|
||||
defaultTopic string
|
||||
}
|
||||
|
||||
func (e engineWithNotifications) Put(o *objectSDK.Object) error {
|
||||
if err := engine.Put(e.e, o); err != nil {
|
||||
if err := e.base.Put(o); err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
|
@ -582,8 +583,16 @@ func (e engineWithNotifications) Put(o *objectSDK.Object) error {
|
|||
|
||||
type engineWithoutNotifications struct {
|
||||
e *engine.StorageEngine
|
||||
|
||||
state util.NodeState
|
||||
}
|
||||
|
||||
func (e engineWithoutNotifications) Put(o *objectSDK.Object) error {
|
||||
if e.state.IsMaintenance() {
|
||||
var st apistatus.NodeUnderMaintenance
|
||||
|
||||
return st
|
||||
}
|
||||
|
||||
return engine.Put(e.e, o)
|
||||
}
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue