Skip to content

Commit

Permalink
feat: filter tablet map using valid candidates before reparenting rep…
Browse files Browse the repository at this point in the history 
…licas

Signed-off-by: Manan Gupta <[email protected]>
  • Loading branch information
@GuptaManan100
GuptaManan100 committed Mar 21, 2024
1 parent ae066f6 commit 126e99b
Show file tree
Hide file tree
Showing 2 changed files with 188 additions and 3 deletions.
11 changes: 9 additions & 2 deletions go/vt/vtctl/reparentutil/emergency_reparenter.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -449,9 +449,16 @@ func (erp *EmergencyReparenter) promoteIntermediateSource(
validCandidateTablets []*topodatapb.Tablet,
opts EmergencyReparentOptions,
) ([]*topodatapb.Tablet, error) {
// we reparent all the other tablets to start replication from our new source
// Create a tablet map from all the valid replicas
validTabletMap := map[string]*topo.TabletInfo{}
for _, candidate := range validCandidateTablets {
alias := topoproto.TabletAliasString(candidate.Alias)
validTabletMap[alias] = tabletMap[alias]
}

// we reparent all the other valid tablets to start replication from our new source
// we wait for all the replicas so that we can choose a better candidate from the ones that started replication later
reachableTablets, err := erp.reparentReplicas(ctx, ev, source, tabletMap, statusMap, opts, true /* waitForAllReplicas */, false /* populateReparentJournal */)
reachableTablets, err := erp.reparentReplicas(ctx, ev, source, validTabletMap, statusMap, opts, true /* waitForAllReplicas */, false /* populateReparentJournal */)
if err != nil {
return nil, err
}
Expand Down
180 changes: 179 additions & 1 deletion go/vt/vtctl/reparentutil/emergency_reparenter_test.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3715,6 +3715,184 @@ func TestEmergencyReparenter_promoteIntermediateSource(t *testing.T) {
},
Hostname: "requires force start",
},
{
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 404,
},
},
},
},
{
name: "success - filter with valid tablets before",
emergencyReparentOps: EmergencyReparentOptions{},
tmc: &testutil.TabletManagerClient{
PopulateReparentJournalResults: map[string]error{
"zone1-0000000100": nil,
},
PrimaryPositionResults: map[string]struct {
Position string
Error error
}{
"zone1-0000000100": {
Error: nil,
},
},
SetReplicationSourceResults: map[string]error{
"zone1-0000000101": nil,
},
},
newSourceTabletAlias: "zone1-0000000100",
tabletMap: map[string]*topo.TabletInfo{
"zone1-0000000100": {
Tablet: &topodatapb.Tablet{
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 100,
},
Hostname: "primary-elect",
},
},
"zone1-0000000101": {
Tablet: &topodatapb.Tablet{
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 101,
},
},
},
"zone1-0000000102": {
Tablet: &topodatapb.Tablet{
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 102,
},
Hostname: "requires force start",
},
},
"zone1-0000000404": {
Tablet: &topodatapb.Tablet{
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 404,
},
Hostname: "ignored tablet",
},
},
},
statusMap: map[string]*replicationdatapb.StopReplicationStatus{
"zone1-0000000101": { // forceStart = false
Before: &replicationdatapb.Status{
IoState: int32(replication.ReplicationStateStopped),
SqlState: int32(replication.ReplicationStateStopped),
},
},
},
keyspace: "testkeyspace",
shard: "-",
shouldErr: false,
result: []*topodatapb.Tablet{
{
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 100,
},
Hostname: "primary-elect",
}, {
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 101,
},
},
},
validCandidateTablets: []*topodatapb.Tablet{
{
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 100,
},
Hostname: "primary-elect",
}, {
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 101,
},
},
},
}, {
name: "success - only 2 tablets and they error",
emergencyReparentOps: EmergencyReparentOptions{},
tmc: &testutil.TabletManagerClient{
PopulateReparentJournalResults: map[string]error{
"zone1-0000000100": nil,
},
PrimaryPositionResults: map[string]struct {
Position string
Error error
}{
"zone1-0000000100": {
Error: nil,
},
},
SetReplicationSourceResults: map[string]error{
"zone1-0000000101": fmt.Errorf("An error"),
},
},
newSourceTabletAlias: "zone1-0000000100",
tabletMap: map[string]*topo.TabletInfo{
"zone1-0000000100": {
Tablet: &topodatapb.Tablet{
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 100,
},
Hostname: "primary-elect",
},
},
"zone1-0000000101": {
Tablet: &topodatapb.Tablet{
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 101,
},
},
},
"zone1-0000000102": {
Tablet: &topodatapb.Tablet{
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 102,
},
Hostname: "requires force start",
},
},
},
statusMap: map[string]*replicationdatapb.StopReplicationStatus{},
keyspace: "testkeyspace",
shard: "-",
shouldErr: false,
result: []*topodatapb.Tablet{
{
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 100,
},
Hostname: "primary-elect",
},
},
validCandidateTablets: []*topodatapb.Tablet{
{
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 100,
},
Hostname: "primary-elect",
}, {
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Uid: 101,
},
},
},
},
{
Expand Down Expand Up @@ -3757,7 +3935,7 @@ func TestEmergencyReparenter_promoteIntermediateSource(t *testing.T) {
},
},
},
"zone1-00000000102": {
"zone1-0000000102": {
Tablet: &topodatapb.Tablet{
Alias: &topodatapb.TabletAlias{
Cell: "zone1",
Expand Down

0 comments on commit 126e99b

Please sign in to comment.