Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

replication mode: fix wrong available store list #7222

Merged
merged 4 commits into from
Oct 18, 2023
Merged

replication mode: fix wrong available store list #7222

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
7f4140c
replication mode: fix wrong available store list
disksing Oct 18, 2023
493bbec
fix build
disksing Oct 18, 2023
d80a4f8
fix test
disksing Oct 18, 2023
ed860a2
address comments
disksing Oct 18, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 1 addition & 5 deletions pkg/replication/replication_mode.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -489,7 +489,7 @@ func (m *ModeManager) tickUpdateState() {
m.drSwitchToSyncRecover()
break
}
if !reflect.DeepEqual(m.drGetAvailableStores(), stores[primaryUp]) && m.drCheckStoreStateUpdated(storeIDs(stores[primaryUp])) {
if !reflect.DeepEqual(m.drGetAvailableStores(), storeIDs(stores[primaryUp])) && m.drCheckStoreStateUpdated(storeIDs(stores[primaryUp])) {
disksing marked this conversation as resolved.
Show resolved Hide resolved
m.drSwitchToAsync(storeIDs(stores[primaryUp]))
}
case drStateSyncRecover:
Expand Down Expand Up @@ -577,10 +577,6 @@ func (m *ModeManager) checkStoreStatus() [][]*core.StoreInfo {
if s.IsRemoved() {
continue
}
// learner peers do not participate in major commit or vote, so it should not count in primary/dr as a normal store.
if s.GetRegionCount() == s.GetLearnerCount() {
continue
}
down := s.DownTime() >= m.config.DRAutoSync.WaitStoreTimeout.Duration
labelValue := s.GetLabelValue(m.config.DRAutoSync.LabelKey)
if labelValue == m.config.DRAutoSync.Primary {
Expand Down
59 changes: 51 additions & 8 deletions pkg/replication/replication_mode_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -245,8 +245,8 @@ func TestStateSwitch(t *testing.T) {
rep.tickUpdateState()
re.Equal(drStateSync, rep.drGetState())

// once the voter node down, even learner node up, swith to async state.
setStoreState(cluster, "up", "up", "up", "up", "down", "up")
// once zone2 down, swith to async state.
setStoreState(cluster, "up", "up", "up", "up", "down", "down")
rep.tickUpdateState()
re.Equal(drStateAsyncWait, rep.drGetState())

Expand All @@ -264,18 +264,18 @@ func TestStateSwitch(t *testing.T) {
re.False(rep.GetReplicationStatus().GetDrAutoSync().GetPauseRegionSplit())

// async_wait -> async_wait
setStoreState(cluster, "up", "up", "up", "up", "down", "up")
setStoreState(cluster, "up", "up", "up", "up", "down", "down")
rep.tickUpdateState()
re.Equal(drStateAsyncWait, rep.drGetState())
assertStateIDUpdate()
rep.tickReplicateStatus()
re.Equal(fmt.Sprintf(`{"state":"async_wait","state_id":%d,"available_stores":[1,2,3,4]}`, stateID), replicator.lastData[1])
setStoreState(cluster, "down", "up", "up", "up", "down", "up")
setStoreState(cluster, "down", "up", "up", "up", "down", "down")
rep.tickUpdateState()
assertStateIDUpdate()
rep.tickReplicateStatus()
re.Equal(fmt.Sprintf(`{"state":"async_wait","state_id":%d,"available_stores":[2,3,4]}`, stateID), replicator.lastData[1])
setStoreState(cluster, "up", "down", "up", "up", "down", "up")
setStoreState(cluster, "up", "down", "up", "up", "down", "down")
rep.tickUpdateState()
assertStateIDUpdate()
rep.tickReplicateStatus()
Expand All @@ -294,7 +294,7 @@ func TestStateSwitch(t *testing.T) {
re.Equal(fmt.Sprintf(`{"state":"async","state_id":%d,"available_stores":[1,3,4]}`, stateID), replicator.lastData[1])

// async -> async
setStoreState(cluster, "up", "up", "up", "up", "down", "up")
setStoreState(cluster, "up", "up", "up", "up", "down", "down")
rep.tickUpdateState()
// store 2 won't be available before it syncs status.
rep.tickReplicateStatus()
Expand All @@ -319,14 +319,14 @@ func TestStateSwitch(t *testing.T) {
// sync_recover -> async
rep.tickUpdateState()
re.Equal(drStateSyncRecover, rep.drGetState())
setStoreState(cluster, "up", "up", "up", "up", "down", "up")
setStoreState(cluster, "up", "up", "up", "up", "down", "down")
rep.tickUpdateState()
re.Equal(drStateAsync, rep.drGetState())
assertStateIDUpdate()
// lost majority, does not switch to async.
rep.drSwitchToSyncRecover()
assertStateIDUpdate()
setStoreState(cluster, "down", "down", "up", "up", "down", "up")
setStoreState(cluster, "down", "down", "up", "up", "down", "down")
rep.tickUpdateState()
re.Equal(drStateSyncRecover, rep.drGetState())

Expand Down Expand Up @@ -636,6 +636,8 @@ func TestComplexPlacementRules(t *testing.T) {
setStoreState(cluster, "up", "up", "up", "up", "up", "up", "up", "down", "up", "down")
rep.tickUpdateState()
re.Equal(drStateAsyncWait, rep.drGetState())
rep.tickReplicateStatus()
re.Equal(fmt.Sprintf(`{"state":"async_wait","state_id":%d,"available_stores":[1,2,3,4,5,6]}`, rep.drAutoSync.StateID), replicator.lastData[1])

// reset to sync
setStoreState(cluster, "up", "up", "up", "up", "up", "up", "up", "up", "up", "up")
Expand Down Expand Up @@ -695,6 +697,47 @@ func TestComplexPlacementRules2(t *testing.T) {
setStoreState(cluster, "up", "up", "up", "up", "down", "down", "up")
rep.tickUpdateState()
re.Equal(drStateAsyncWait, rep.drGetState())
rep.tickReplicateStatus()
re.Equal(fmt.Sprintf(`{"state":"async_wait","state_id":%d,"available_stores":[1,2,3,4]}`, rep.drAutoSync.StateID), replicator.lastData[1])
}

func TestComplexPlacementRules3(t *testing.T) {
re := require.New(t)
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
store := storage.NewStorageWithMemoryBackend()
conf := config.ReplicationModeConfig{ReplicationMode: modeDRAutoSync, DRAutoSync: config.DRAutoSyncReplicationConfig{
LabelKey: "zone",
Primary: "zone1",
DR: "zone2",
WaitStoreTimeout: typeutil.Duration{Duration: time.Minute},
}}
cluster := mockcluster.NewCluster(ctx, mockconfig.NewTestOptions())
replicator := newMockReplicator([]uint64{1})
rep, err := NewReplicationModeManager(conf, store, cluster, replicator)
re.NoError(err)
cluster.GetRuleManager().SetAllGroupBundles(
genPlacementRuleConfig([]ruleConfig{
{key: "logic", value: "logic1", role: placement.Voter, count: 2},
{key: "logic", value: "logic2", role: placement.Learner, count: 1},
{key: "logic", value: "logic3", role: placement.Voter, count: 1},
}), true)

cluster.AddLabelsStore(1, 1, map[string]string{"zone": "zone1", "logic": "logic1"})
cluster.AddLabelsStore(2, 1, map[string]string{"zone": "zone1", "logic": "logic1"})
cluster.AddLabelsStore(3, 1, map[string]string{"zone": "zone1", "logic": "logic2"})
cluster.AddLabelsStore(4, 1, map[string]string{"zone": "zone1", "logic": "logic2"})
cluster.AddLabelsStore(5, 1, map[string]string{"zone": "zone2", "logic": "logic3"})

// initial state is sync
re.Equal(drStateSync, rep.drGetState())

// zone2 down, switch state, available stores should contain logic2 (learner)
setStoreState(cluster, "up", "up", "up", "up", "down")
rep.tickUpdateState()
re.Equal(drStateAsyncWait, rep.drGetState())
rep.tickReplicateStatus()
re.Equal(fmt.Sprintf(`{"state":"async_wait","state_id":%d,"available_stores":[1,2,3,4]}`, rep.drAutoSync.StateID), replicator.lastData[1])
}

func genRegions(cluster *mockcluster.Cluster, stateID uint64, n int) []*core.RegionInfo {
Expand Down
Loading