Skip to content

hc remove time difference public api mention #6335

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
StekPerepolnen merged 1 commit into from
Jul 15, 2024
Merged

hc remove time difference public api mention #6335

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
40 changes: 12 additions & 28 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 @@ -242,6 +242,7 @@ class TSelfCheckRequest : public TActorBootstrapped<TSelfCheckRequest> {
ui64 StorageQuota = 0;
ui64 StorageUsage = 0;
TMaybeServerlessComputeResourcesMode ServerlessComputeResourcesMode;
TNodeId MaxTimeDifferenceNodeId = 0;
TString Path;
};

Expand Down Expand Up @@ -566,20 +567,6 @@ class TSelfCheckRequest : public TActorBootstrapped<TSelfCheckRequest> {
return FilterDatabase && FilterDatabase != DomainPath;
}

bool IsTimeDifferenceCheckNode(const TNodeId nodeId) const {
if (!IsSpecificDatabaseFilter()) {
return true;
}

auto it = DatabaseState.find(FilterDatabase);
if (it == DatabaseState.end()) {
return false;
}
auto& computeNodeIds = it->second.ComputeNodeIds;

return std::find(computeNodeIds.begin(), computeNodeIds.end(), nodeId) != computeNodeIds.end();
}

void Bootstrap() {
FilterDatabase = Request->Database;
if (Request->Request.operation_params().has_operation_timeout()) {
Expand Down Expand Up @@ -1450,7 +1437,7 @@ class TSelfCheckRequest : public TActorBootstrapped<TSelfCheckRequest> {
}
}

void FillComputeNodeStatus(TDatabaseState& databaseState, TNodeId nodeId, Ydb::Monitoring::ComputeNodeStatus& computeNodeStatus, TSelfCheckContext context, bool reportTimeDifference) {
void FillComputeNodeStatus(TDatabaseState& databaseState, TNodeId nodeId, Ydb::Monitoring::ComputeNodeStatus& computeNodeStatus, TSelfCheckContext context) {
FillNodeInfo(nodeId, context.Location.mutable_compute()->mutable_node());

TSelfCheckContext rrContext(&context, "NODE_UPTIME");
Expand Down Expand Up @@ -1502,17 +1489,15 @@ class TSelfCheckRequest : public TActorBootstrapped<TSelfCheckRequest> {
status = Ydb::Monitoring::StatusFlag::GREEN;
}

computeNodeStatus.mutable_max_time_difference()->set_peer(ToString(peerId));
computeNodeStatus.mutable_max_time_difference()->set_difference_ms(timeDifferenceDuration.MilliSeconds());
computeNodeStatus.set_overall(status);

if (reportTimeDifference) {
if (databaseState.MaxTimeDifferenceNodeId == nodeId) {
TSelfCheckContext tdContext(&context, "NODES_TIME_DIFFERENCE");
FillNodeInfo(peerId, tdContext.Location.mutable_compute()->mutable_peer());
if (status == Ydb::Monitoring::StatusFlag::GREEN) {
tdContext.ReportStatus(status);
} else {
tdContext.ReportStatus(status, TStringBuilder() << "The nodes have a time difference of " << timeDifferenceDuration.MilliSeconds() << " ms", ETags::SyncState);
tdContext.ReportStatus(status, TStringBuilder() << "Node is "
<< timeDifferenceDuration.MilliSeconds() << " ms "
<< (timeDifferenceUs > 0 ? "behind " : "ahead of ")
<< "peer [" << peerId << "]", ETags::SyncState);
}
}
}
Expand Down Expand Up @@ -1580,21 +1565,20 @@ class TSelfCheckRequest : public TActorBootstrapped<TSelfCheckRequest> {
if (systemStatus != Ydb::Monitoring::StatusFlag::GREEN && systemStatus != Ydb::Monitoring::StatusFlag::GREY) {
context.ReportStatus(systemStatus, "Compute has issues with system tablets", ETags::ComputeState, {ETags::SystemTabletState});
}
long maxClockSkewUs = 0;
TNodeId maxClockSkewNodeId = 0;
long maxTimeDifferenceUs = 0;
for (TNodeId nodeId : *computeNodeIds) {
auto itNodeSystemState = MergedNodeSystemState.find(nodeId);
if (itNodeSystemState != MergedNodeSystemState.end()) {
if (std::count(computeNodeIds->begin(), computeNodeIds->end(), itNodeSystemState->second->GetMaxClockSkewPeerId()) > 0
&& abs(itNodeSystemState->second->GetMaxClockSkewWithPeerUs()) > maxClockSkewUs) {
maxClockSkewUs = abs(itNodeSystemState->second->GetMaxClockSkewWithPeerUs());
maxClockSkewNodeId = nodeId;
&& abs(itNodeSystemState->second->GetMaxClockSkewWithPeerUs()) > maxTimeDifferenceUs) {
maxTimeDifferenceUs = abs(itNodeSystemState->second->GetMaxClockSkewWithPeerUs());
databaseState.MaxTimeDifferenceNodeId = nodeId;
}
}
}
for (TNodeId nodeId : *computeNodeIds) {
auto& computeNode = *computeStatus.add_nodes();
FillComputeNodeStatus(databaseState, nodeId, computeNode, {&context, "COMPUTE_NODE"}, maxClockSkewNodeId == nodeId);
FillComputeNodeStatus(databaseState, nodeId, computeNode, {&context, "COMPUTE_NODE"});
}
FillComputeDatabaseStatus(databaseState, computeStatus, {&context, "COMPUTE_QUOTA"});
context.ReportWithMaxChildStatus("Some nodes are restarting too often", ETags::ComputeState, {ETags::Uptime});
Expand Down
8 changes: 0 additions & 8 deletions ydb/public/api/protos/ydb_monitoring.proto
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -106,19 +106,12 @@ message LoadAverageStatus {
uint32 cores = 3;
}

message TimeDifferenceStatus {
StatusFlag.Status overall = 1;
int64 difference_ms = 2;
string peer = 3;
}

message ComputeNodeStatus {
string id = 1;
StatusFlag.Status overall = 2;
repeated ComputeTabletStatus tablets = 3;
repeated ThreadPoolStatus pools = 4;
LoadAverageStatus load = 5;
TimeDifferenceStatus max_time_difference = 6;
}

message ComputeStatus {
Expand Down Expand Up @@ -174,7 +167,6 @@ message LocationCompute {
LocationNode node = 1;
LocationComputePool pool = 2;
LocationComputeTablet tablet = 3;
LocationNode peer = 4;
}

message LocationDatabase {
Expand Down