Skip to content

do not let faulty pdisks make group status dead in healthcheck #9744

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
merged 1 commit into from
Oct 16, 2024
Merged

do not let faulty pdisks make group status dead in healthcheck #9744

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 2 additions & 0 deletions ydb/core/health_check/health_check.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1968,6 +1968,8 @@ class TSelfCheckRequest : public TActorBootstrapped<TSelfCheckRequest> {
}
}

// do not propagate RED status to vdisk - so that vdisk is not considered down when computing group status
context.OverallStatus = MinStatus(context.OverallStatus, Ydb::Monitoring::StatusFlag::ORANGE);
storagePDiskStatus.set_overall(context.GetOverallStatus());
}

Expand Down
25 changes: 20 additions & 5 deletions ydb/core/health_check/health_check_ut.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,7 +68,8 @@ Y_UNIT_TEST_SUITE(THealthCheckTest) {

struct TTestVSlotInfo {
std::optional<NKikimrBlobStorage::EVDiskStatus> Status;
ui32 Generation;
ui32 Generation = DEFAULT_GROUP_GENERATION;
NKikimrBlobStorage::EDriveStatus PDiskStatus = NKikimrBlobStorage::ACTIVE;

TTestVSlotInfo(std::optional<NKikimrBlobStorage::EVDiskStatus> status = NKikimrBlobStorage::READY,
ui32 generation = DEFAULT_GROUP_GENERATION)
Expand All @@ -77,7 +78,11 @@ Y_UNIT_TEST_SUITE(THealthCheckTest) {
{
}

TTestVSlotInfo(NKikimrBlobStorage::EVDiskStatus status) : Status(status), Generation(DEFAULT_GROUP_GENERATION) {}
TTestVSlotInfo(NKikimrBlobStorage::EVDiskStatus status, NKikimrBlobStorage::EDriveStatus pDiskStatus = NKikimrBlobStorage::ACTIVE)
: Status(status)
, PDiskStatus(pDiskStatus)
{
}
};

using TVDisks = TVector<TTestVSlotInfo>;
Expand Down Expand Up @@ -222,18 +227,20 @@ Y_UNIT_TEST_SUITE(THealthCheckTest) {
entry->mutable_info()->set_name(STORAGE_POOL_NAME);
}

void AddPDisksToSysViewResponse(NSysView::TEvSysView::TEvGetPDisksResponse::TPtr* ev, size_t count, double occupancy) {
void AddPDisksToSysViewResponse(NSysView::TEvSysView::TEvGetPDisksResponse::TPtr* ev, const TVDisks& vslots, double occupancy) {
auto& record = (*ev)->Get()->Record;
auto entrySample = record.entries(0);
record.clear_entries();
auto pdiskId = PDISK_START_ID;
const size_t totalSize = 3'200'000'000'000ull;
for (size_t i = 0; i < count; ++i) {
const auto *descriptor = NKikimrBlobStorage::EDriveStatus_descriptor();
for (const auto& vslot : vslots) {
auto* entry = record.add_entries();
entry->CopyFrom(entrySample);
entry->mutable_key()->set_pdiskid(pdiskId);
entry->mutable_info()->set_totalsize(totalSize);
entry->mutable_info()->set_availablesize((1 - occupancy) * totalSize);
entry->mutable_info()->set_statusv2(descriptor->FindValueByNumber(vslot.PDiskStatus)->name());
++pdiskId;
}
}
Expand Down Expand Up @@ -482,7 +489,7 @@ Y_UNIT_TEST_SUITE(THealthCheckTest) {
}
case NSysView::TEvSysView::EvGetPDisksResponse: {
auto* x = reinterpret_cast<NSysView::TEvSysView::TEvGetPDisksResponse::TPtr*>(&ev);
AddPDisksToSysViewResponse(x, vdisks.size(), occupancy);
AddPDisksToSysViewResponse(x, vdisks, occupancy);
break;
}
case NSysView::TEvSysView::EvGetGroupsResponse: {
Expand Down Expand Up @@ -710,6 +717,14 @@ Y_UNIT_TEST_SUITE(THealthCheckTest) {
CheckHcResultHasIssuesWithStatus(result, "STORAGE_GROUP", Ydb::Monitoring::StatusFlag::RED, 1);
}

Y_UNIT_TEST(YellowIssueReadyVDisksOnFaultyPDisks) {
auto result = RequestHcWithVdisks(NKikimrBlobStorage::TGroupStatus::PARTIAL, TVDisks{3, {NKikimrBlobStorage::READY, NKikimrBlobStorage::FAULTY}});
Cerr << result.ShortDebugString() << Endl;
CheckHcResultHasIssuesWithStatus(result, "STORAGE_GROUP", Ydb::Monitoring::StatusFlag::YELLOW, 1);
CheckHcResultHasIssuesWithStatus(result, "STORAGE_GROUP", Ydb::Monitoring::StatusFlag::ORANGE, 0);
CheckHcResultHasIssuesWithStatus(result, "STORAGE_GROUP", Ydb::Monitoring::StatusFlag::RED, 0);
}

/* HC currently infers group status on its own, so it's never unknown
Y_UNIT_TEST(RedGroupIssueWhenUnknownGroupStatus) {
auto result = RequestHcWithVdisks(NKikimrBlobStorage::TGroupStatus::UNKNOWN, {});
Expand Down
Loading