Skip to content

Limit SelfHeal reassign requests in-flight #17618

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

Merged
serbel324 merged 6 commits into from
Apr 25, 2025
Merged

Limit SelfHeal reassign requests in-flight #17618

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
661aa7b
Limit in flight self-heal reassign requests by 1
serbel324 Apr 23, 2025
a0981bd
Don't allow GroupLayoutSanitizer to create ReassignerActor if there i…
serbel324 Apr 23, 2025
2b62f21
Fix stuck reassigns
serbel324 Apr 24, 2025
689ccb6
Check Reassigner ActorId
serbel324 Apr 25, 2025
4709073
Remove copypaste
serbel324 Apr 25, 2025
ce0c35f
Don't enqueue unreassignable groups for reassign
serbel324 Apr 25, 2025
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
67 changes: 67 additions & 0 deletions ydb/core/blobstorage/ut_blobstorage/self_heal.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,67 @@
#include <ydb/core/blobstorage/ut_blobstorage/lib/env.h>

Y_UNIT_TEST_SUITE(SelfHeal) {
void TestReassignThrottling() {
const TBlobStorageGroupType erasure = TBlobStorageGroupType::ErasureMirror3dc;
const ui32 groupsCount = 32;

TEnvironmentSetup env({
.NodeCount = erasure.BlobSubgroupSize(),
.Erasure = erasure,
});

// create 2 pdisks per node to allow self-healings and
// allocate groups
env.CreateBoxAndPool(2, groupsCount);
env.Sim(TDuration::Minutes(1));

auto base = env.FetchBaseConfig();
UNIT_ASSERT_VALUES_EQUAL(base.GroupSize(), groupsCount);

ui32 maxReassignsInFlight = 0;

std::set<TActorId> reassignersInFlight;

auto catchReassigns = [&](ui32 /*nodeId*/, std::unique_ptr<IEventHandle>& ev) {
if (ev->GetTypeRewrite() == TEvBlobStorage::TEvControllerConfigRequest::EventType) {
const auto& request = ev->Get<TEvBlobStorage::TEvControllerConfigRequest>()->Record.GetRequest();
for (const auto& command : request.GetCommand()) {
if (command.GetCommandCase() == NKikimrBlobStorage::TConfigRequest::TCommand::kReassignGroupDisk) {
UNIT_ASSERT(!reassignersInFlight.contains(ev->Sender));
reassignersInFlight.insert(ev->Sender);
maxReassignsInFlight = std::max(maxReassignsInFlight, (ui32)reassignersInFlight.size());
}
}
} else if (ev->GetTypeRewrite() == TEvBlobStorage::TEvControllerConfigResponse::EventType) {
auto it = reassignersInFlight.find(ev->Recipient);
if (it != reassignersInFlight.end()) {
reassignersInFlight.erase(it);
}
}
return true;
};

env.Runtime->FilterFunction = catchReassigns;

auto pdisk = base.GetPDisk(0);
// set FAULTY status on the chosen PDisk
{
NKikimrBlobStorage::TConfigRequest request;
auto* cmd = request.AddCommand()->MutableUpdateDriveStatus();
cmd->MutableHostKey()->SetNodeId(pdisk.GetNodeId());
cmd->SetPDiskId(pdisk.GetPDiskId());
cmd->SetStatus(NKikimrBlobStorage::FAULTY);
auto res = env.Invoke(request);
UNIT_ASSERT_C(res.GetSuccess(), res.GetErrorDescription());
UNIT_ASSERT_C(res.GetStatus(0).GetSuccess(), res.GetStatus(0).GetErrorDescription());
}

env.Sim(TDuration::Minutes(15));

UNIT_ASSERT_C(maxReassignsInFlight == 1, "maxReassignsInFlight# " << maxReassignsInFlight);
}

Y_UNIT_TEST(ReassignThrottling) {
TestReassignThrottling();
}
}
1 change: 1 addition & 0 deletions ydb/core/blobstorage/ut_blobstorage/ya.make
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,6 +41,7 @@ SRCS(
recovery.cpp
sanitize_groups.cpp
scrub_fast.cpp
self_heal.cpp
shred.cpp
snapshots.cpp
space_check.cpp
Expand Down
179 changes: 129 additions & 50 deletions ydb/core/mind/bscontroller/self_heal.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -253,13 +253,19 @@ namespace NKikimr::NBsController {
struct TWithFaultyDisks {};
struct TWithInvalidLayout {};

enum class EReassignStatus : ui8 {
NotNeeded = 0,
Enqueued,
Active,
};

struct TGroupRecord
: TIntrusiveListItem<TGroupRecord, TWithFaultyDisks>
, TIntrusiveListItem<TGroupRecord, TWithInvalidLayout>
{
const TGroupId GroupId;
TEvControllerUpdateSelfHealInfo::TGroupContent Content;
TActorId ReassignerActorId; // reassigner in flight
EReassignStatus ReassignStatus = EReassignStatus::NotNeeded;
TDuration RetryTimeout = MinRetryTimeout;
TMonotonic NextRetryTimestamp = TMonotonic::Zero();
std::shared_ptr<TBlobStorageGroupInfo::TTopology> Topology;
Expand All @@ -278,7 +284,8 @@ namespace NKikimr::NBsController {
THashMap<TGroupId, TGroupRecord> Groups;
TIntrusiveList<TGroupRecord, TWithFaultyDisks> GroupsWithFaultyDisks;
TIntrusiveList<TGroupRecord, TWithInvalidLayout> GroupsWithInvalidLayout;
std::shared_ptr<std::atomic_uint64_t> UnreassignableGroups;
std::unordered_set<TGroupId> UnreassignableGroups;
std::shared_ptr<std::atomic_uint64_t> UnreassignableGroupsCount;
bool GroupLayoutSanitizerEnabled;
bool AllowMultipleRealmsOccupation;
bool DonorMode;
Expand All @@ -294,13 +301,17 @@ namespace NKikimr::NBsController {
static constexpr uint32_t GroupLayoutSanitizerOperationLogSize = 128;
TOperationLog<GroupLayoutSanitizerOperationLogSize> GroupLayoutSanitizerOperationLog;

std::deque<TGroupId> SelfHealReassignQueue;
std::deque<TGroupId> GroupLayoutSanitizerReassignQueue;
std::optional<TActorId> ActiveReassignerActorId = std::nullopt;

public:
TSelfHealActor(ui64 tabletId, std::shared_ptr<std::atomic_uint64_t> unreassignableGroups, THostRecordMap hostRecords,
bool groupLayoutSanitizerEnabled, bool allowMultipleRealmsOccupation, bool donorMode,
std::shared_ptr<TControlWrapper> enableSelfHealWithDegraded,
std::shared_ptr<std::atomic_uint64_t> groupsWithInvalidLayoutCounter)
: TabletId(tabletId)
, UnreassignableGroups(std::move(unreassignableGroups))
, UnreassignableGroupsCount(std::move(unreassignableGroups))
, GroupLayoutSanitizerEnabled(groupLayoutSanitizerEnabled)
, AllowMultipleRealmsOccupation(allowMultipleRealmsOccupation)
, DonorMode(donorMode)
Expand Down Expand Up @@ -385,8 +396,11 @@ namespace NKikimr::NBsController {
TGroupRecord& group = it->second;

// kill reassigner, if it is working
if (group.ReassignerActorId) {
Send(group.ReassignerActorId, new TEvents::TEvPoison);
if (group.ReassignStatus == EReassignStatus::Active) {
Y_DEBUG_ABORT_UNLESS(ActiveReassignerActorId);
if (ActiveReassignerActorId) {
Send(*ActiveReassignerActorId, new TEvents::TEvPoison);
}
}

// remove the group
Expand Down Expand Up @@ -422,49 +436,16 @@ namespace NKikimr::NBsController {
void CheckGroups() {
const TMonotonic now = TActivationContext::Monotonic();

ui64 counter = 0;

for (TGroupRecord& group : GroupsWithFaultyDisks) {
if (group.ReassignerActorId || now < group.NextRetryTimestamp) {
continue; // we are already running reassigner for this group
if (group.ReassignStatus != EReassignStatus::NotNeeded || now < group.NextRetryTimestamp) {
continue; // reassign is already enqueued
}

if (group.UpdateConfigTxSeqNo < group.ResponseConfigTxSeqNo) {
continue; // response from bsc was received before selfheal info update
}

// check if it is possible to move anything out
bool isSelfHealReasonDecommit;
bool ignoreDegradedGroupsChecks;
if (const auto v = FindVDiskToReplace(group.Content, now, group.Topology.get(), &isSelfHealReasonDecommit,
&ignoreDegradedGroupsChecks)) {
group.ReassignerActorId = Register(new TReassignerActor(ControllerId, group.GroupId, group.Content,
*v, group.Topology, isSelfHealReasonDecommit, ignoreDegradedGroupsChecks, DonorMode));
} else {
++counter; // this group can't be reassigned right now

auto log = [&]() {
TStringStream ss;
ss << "[";
bool first = true;
for (const auto& [vdiskId, vdisk] : group.Content.VDisks) {
if (!std::exchange(first, false)) {
ss << ",";
}
ss << "{";
ss << vdiskId;
ss << (IsReady(vdisk, now) ? " Ready" : " NotReady");
ss << (vdisk.Faulty ? " Faulty" : "");
ss << (vdisk.Bad ? " IsBad" : "");
ss << (vdisk.Decommitted ? " Decommitted" : "");
ss << "}";
}
ss << "]";
return ss.Str();
};

STLOG(PRI_INFO, BS_SELFHEAL, BSSH11, "group can't be reassigned right now " << log(), (GroupId, group.GroupId));
}

EnqueueReassign(group, EGroupRepairOperation::SelfHeal);
}

if (GroupLayoutSanitizerEnabled) {
Expand All @@ -488,20 +469,19 @@ namespace NKikimr::NBsController {
}

Y_ABORT_UNLESS(!group.LayoutValid);
if (group.ReassignerActorId || now < group.NextRetryTimestamp) {
if (group.ReassignStatus != EReassignStatus::NotNeeded || now < group.NextRetryTimestamp) {
// nothing to do
} else {
ADD_RECORD_WITH_TIMESTAMP_TO_OPERATION_LOG(GroupLayoutSanitizerOperationLog,
"Start sanitizing GroupId# " << group.GroupId << " GroupGeneration# " << group.Content.Generation);
group.ReassignerActorId = Register(new TReassignerActor(ControllerId, group.GroupId, group.Content,
std::nullopt, group.Topology, false /*isSelfHealReasonDecommit*/,
false /*ignoreDegradedGroupsChecks*/, DonorMode));
EnqueueReassign(group, EGroupRepairOperation::GroupLayoutSanitizer);
}
}
}

ProcessReassignQueues();
GroupsWithInvalidLayoutCounter->store(GroupsWithInvalidLayout.Size());
UnreassignableGroups->store(counter);
UnreassignableGroupsCount->store(UnreassignableGroups.size());
}

void UpdateGroupLayoutInformation(TGroupRecord& group) {
Expand Down Expand Up @@ -602,9 +582,13 @@ namespace NKikimr::NBsController {
}

void Handle(TEvReassignerDone::TPtr& ev) {
if (const auto it = Groups.find(ev->Get()->GroupId); it != Groups.end() && it->second.ReassignerActorId == ev->Sender) {
Y_ABORT_UNLESS(ActiveReassignerActorId);
TActorId reassigner = *std::exchange(ActiveReassignerActorId, std::nullopt);
Y_ABORT_UNLESS(reassigner == ev->Sender);

if (const auto it = Groups.find(ev->Get()->GroupId); it != Groups.end()) {
auto& group = it->second;
group.ReassignerActorId = {};
group.ReassignStatus = EReassignStatus::NotNeeded;

const TMonotonic now = TActivationContext::Monotonic();
if (ev->Get()->Success) {
Expand All @@ -623,9 +607,9 @@ namespace NKikimr::NBsController {
"Sanitizing failed GroupId# " << group.GroupId << " ErrorReason# " << ev->Get()->ErrorReason);
}
}

CheckGroups();
}
ProcessReassignQueues();
}

using TVDiskInfo = TEvControllerUpdateSelfHealInfo::TGroupContent::TVDiskInfo;
Expand Down Expand Up @@ -654,6 +638,101 @@ namespace NKikimr::NBsController {
Send(ev->Sender, new NMon::TEvRemoteHttpInfoRes(str.Str()));
}

void ProcessReassignQueues() {
while (!ActiveReassignerActorId && !SelfHealReassignQueue.empty()) {
TGroupId groupId = SelfHealReassignQueue.front();
SelfHealReassignQueue.pop_front();
CreateReassignerActorIfNeededForSelfHeal(groupId);
}

while (!ActiveReassignerActorId && !GroupLayoutSanitizerReassignQueue.empty()) {
TGroupId groupId = GroupLayoutSanitizerReassignQueue.front();
GroupLayoutSanitizerReassignQueue.pop_front();
auto it = Groups.find(groupId);
if (it != Groups.end()) {
TGroupRecord& group = it->second;
CreateReassignerActor(group, std::nullopt, false, false);
}
}
}

bool CreateReassignerActorIfNeededForSelfHeal(TGroupId groupId) {
auto it = Groups.find(groupId);
if (it == Groups.end()) {
// group is deleted
return false;
}

TGroupRecord& group = it->second;
if (group.ReassignStatus == EReassignStatus::NotNeeded) {
// Group is already fully healed
return false;
}

// check if it is possible to move anything out
bool isSelfHealReasonDecommit;
bool ignoreDegradedGroupsChecks;
if (const std::optional<TVDiskID> vdiskId = FindVDiskToReplace(group.Content, TActivationContext::Monotonic(),
group.Topology.get(), &isSelfHealReasonDecommit, &ignoreDegradedGroupsChecks)) {
if (auto it = UnreassignableGroups.find(groupId); it != UnreassignableGroups.end()) {
UnreassignableGroups.erase(it);
}
CreateReassignerActor(group, vdiskId, isSelfHealReasonDecommit, ignoreDegradedGroupsChecks);
return true;
} else {
// unable to reassign VDisk
UnreassignableGroups.insert(groupId);
group.ReassignStatus = EReassignStatus::NotNeeded;

TMonotonic now = TActivationContext::Monotonic();
auto log = [&]() {
TStringStream ss;
ss << "[";
bool first = true;
for (const auto& [vdiskId, vdisk] : group.Content.VDisks) {
if (!std::exchange(first, false)) {
ss << ",";
}
ss << "{";
ss << vdiskId;
ss << (IsReady(vdisk, now) ? " Ready" : " NotReady");
ss << (vdisk.Faulty ? " Faulty" : "");
ss << (vdisk.Bad ? " IsBad" : "");
ss << (vdisk.Decommitted ? " Decommitted" : "");
ss << "}";
}
ss << "]";
return ss.Str();
};

STLOG(PRI_INFO, BS_SELFHEAL, BSSH11, "group can't be reassigned right now " << log(), (GroupId, groupId));
}
return false;
}


void CreateReassignerActor(TGroupRecord& group, std::optional<TVDiskID> vdiskId, bool isSelfHealReasonDecommit,
bool ignoreDegradedGroupsChecks) {
group.ReassignStatus = EReassignStatus::Active;
Y_ABORT_UNLESS(!ActiveReassignerActorId);
ActiveReassignerActorId = Register(new TReassignerActor(ControllerId, group.GroupId, group.Content,
vdiskId, group.Topology, isSelfHealReasonDecommit, ignoreDegradedGroupsChecks, DonorMode));
}

void EnqueueReassign(TGroupRecord& group, EGroupRepairOperation operation) {
group.ReassignStatus = EReassignStatus::Enqueued;
switch (operation) {
case EGroupRepairOperation::SelfHeal:
SelfHealReassignQueue.push_back(group.GroupId);
break;
case EGroupRepairOperation::GroupLayoutSanitizer:
GroupLayoutSanitizerReassignQueue.push_back(group.GroupId);
break;
default:
Y_ABORT("Unknown operation");
}
}

void RenderMonPage(IOutputStream& out, bool selfHealEnabled) {
HTML(out) {
TAG(TH2) {
Expand Down
Loading