Skip to content

Improve CA introspection #30263

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
PavelEfarinov merged 2 commits into from
Dec 10, 2025
Merged

Improve CA introspection #30263

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a82d79b
Improve CA introspection
Hor911 Dec 6, 2025
95da9e4
Fix UT
Hor911 Dec 7, 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
195 changes: 190 additions & 5 deletions ydb/library/yql/dq/actors/compute/dq_compute_actor_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -406,6 +406,7 @@ class TDqComputeActorBase : public NActors::TActorBootstrapped<TDerived>
}

ProcessOutputsState.LastRunStatus = status;
ProcessOutputsState.LastRunTime = TInstant::Now();

for (auto& entry : OutputChannelsMap) {
const ui64 channelId = entry.first;
Expand Down Expand Up @@ -855,25 +856,29 @@ class TDqComputeActorBase : public NActors::TActorBootstrapped<TDerived>

protected:
struct TInputChannelInfo {
ui32 InputIndex;
TString LogPrefix;
ui64 ChannelId;
ui32 SrcStageId;
IDqInputChannel::TPtr Channel;
bool HasPeer = false;
NActors::TActorId PeerId;
const NDqProto::EWatermarksMode WatermarksMode;
const TDuration WatermarksIdleTimeout = TDuration::Max();
std::optional<NDqProto::TCheckpoint> PendingCheckpoint;
const NDqProto::ECheckpointingMode CheckpointingMode;
i64 FreeSpace = 0;

explicit TInputChannelInfo(
ui32 inputIndex,
const TString& logPrefix,
ui64 channelId,
ui32 srcStageId,
NDqProto::EWatermarksMode watermarksMode,
NDqProto::ECheckpointingMode checkpointingMode,
TDuration watermarksIdleTimeout)
: LogPrefix(logPrefix)
: InputIndex(inputIndex)
, LogPrefix(logPrefix)
, ChannelId(channelId)
, SrcStageId(srcStageId)
, WatermarksMode(watermarksMode)
Expand Down Expand Up @@ -919,6 +924,7 @@ class TDqComputeActorBase : public NActors::TActorBootstrapped<TDerived>
ui32 DstStageId;
IDqOutputChannel::TPtr Channel;
bool HasPeer = false;
NActors::TActorId PeerId;
bool Finished = false; // != Channel->IsFinished() // If channel is in finished state, it sends only checkpoints.
bool EarlyFinish = false;
bool PopStarted = false;
Expand Down Expand Up @@ -1089,6 +1095,7 @@ class TDqComputeActorBase : public NActors::TActorBootstrapped<TDerived>

Channels->SetInputChannelPeer(channelUpdate.GetId(), peer);
inputChannel->HasPeer = true;
inputChannel->PeerId = peer;

continue;
}
Expand All @@ -1101,6 +1108,7 @@ class TDqComputeActorBase : public NActors::TActorBootstrapped<TDerived>

Channels->SetOutputChannelPeer(channelUpdate.GetId(), peer);
outputChannel->HasPeer = true;
outputChannel->PeerId = peer;
if (outputChannel->Channel) {
outputChannel->Channel->UpdateSettings({.IsLocalChannel = peer.NodeId() == this->SelfId().NodeId()});
}
Expand Down Expand Up @@ -1276,8 +1284,7 @@ class TDqComputeActorBase : public NActors::TActorBootstrapped<TDerived>
void MonitoringExtra(TStringStream&) {
}

void OnMonitoringPage(NActors::NMon::TEvHttpInfo::TPtr& ev) {
TStringStream html;
void DumpForMonitoring(TStringStream& html) {
static_cast<TDerived*>(this)->MonitoringExtra(html);

#define DUMP(P, X,...) html << #X ": " << P.X __VA_ARGS__ << "<br />"
Expand Down Expand Up @@ -1533,8 +1540,181 @@ class TDqComputeActorBase : public NActors::TActorBootstrapped<TDerived>
}
#undef DUMP
#undef DUMP_PREFIXED
}

void DefaultMonitoringPage(TStringStream& str) {
HTML(str) {
PRE() {
Copy link
Collaborator

@yumkam yumkam Dec 9, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Не уверен, что всё целиком нужно заворачивать в <pre>

Copy link
Collaborator Author

@Hor911 Hor911 Dec 10, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

У нас в таком стиле много другие страниц внутри YDB сделано, поэтому я оттуда скопировал, но можно обсудить и покрутить, как лучше будет выглядеть

str << "TDqComputeActorBase, SelfId=" << this->SelfId() << ' ';
HREF(TStringBuilder() << "?ca=" << this->SelfId() << "&view=dump") {
str << "Dump";
}
str << ' ';
HREF(TStringBuilder() << "?ca=" << this->SelfId() << "&view=run") {
str << "Run";
}
str << Endl;

COLLAPSED_BUTTON_CONTENT("ProcessOutputsState", TStringBuilder() << "ProcessOutputsState: " << ProcessOutputsState.LastRunTime << ' ' << ProcessOutputsState.LastRunStatus) {
str << " Inflight: " << ProcessOutputsState.Inflight << Endl;
str << " ChannelsReady: " << ProcessOutputsState.ChannelsReady << Endl;
str << " HasDataToSend: " << ProcessOutputsState.HasDataToSend << Endl;
str << " DataWasSent: " << ProcessOutputsState.DataWasSent << Endl;
str << " AllOutputsFinished: " << ProcessOutputsState.AllOutputsFinished << Endl;
str << " LastRunStatus: " << ProcessOutputsState.LastRunStatus << Endl;
str << " LastRunTime: " << ProcessOutputsState.LastRunTime << Endl;
str << " LastPopReturnedNoData: " << ProcessOutputsState.LastPopReturnedNoData << Endl;
}

str << Endl << "Input Channels:" << Endl;
TABLE_SORTABLE_CLASS("table table-condensed") {
TABLEHEAD() {
TABLER() {
TABLEH_ATTRS({{"title", "ChannelId"}}) {str << "Id";}
TABLEH_ATTRS({{"title", "SrcStageId"}}) {str << "Src";}
TABLEH_ATTRS({{"title", "InputIndex"}}) {str << "Idx";}
TABLEH() {str << "PeerId";}
TABLEH_ATTRS({{"title", "IsFinished"}}) {str << "F";}
TABLEH() {str << "Push.Bytes";}
TABLEH() {str << "Push.Rows";}
TABLEH() {str << "Pop.Bytes";}
TABLEH() {str << "Pop.Rows";}
}
}
TABLEBODY() {
for (const auto& [id, info]: InputChannelsMap) {
TABLER() {
TABLED() {str << info.ChannelId;}
TABLED() {str << info.SrcStageId;}
TABLED() {str << info.InputIndex;}
TABLED() {
if (info.HasPeer) {
HREF(TStringBuilder() << "/node/" << info.PeerId.NodeId() << "/actors/kqp_node?ca=" << info.PeerId) {
str << info.PeerId;
}
} else {
str << "N/A";
}
}

this->Send(ev->Sender, new NActors::NMon::TEvHttpInfoRes(html.Str()));
auto channel = info.Channel;
if (!channel) {
auto stats = GetTaskRunnerStats();
if (stats) {
auto stageIt = stats->InputChannels.find(info.SrcStageId);
if (stageIt != stats->InputChannels.end()) {
auto channelIt = stageIt->second.find(info.ChannelId);
if (channelIt != stageIt->second.end()) {
channel = channelIt->second;
}
}
}
}

if (channel) {
TABLED() {str << channel->IsFinished();}
auto& pushStats = channel->GetPushStats();
TABLED() {str << pushStats.Bytes;}
TABLED() {str << pushStats.Rows;}
auto& popStats = channel->GetPopStats();
TABLED() {str << popStats.Bytes;}
TABLED() {str << popStats.Rows;}
} else {
TABLED() {str << "N/A";}
Copy link
Collaborator

@yumkam yumkam Dec 9, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: TABLED_ATTRS({{"colspan,"5"}})? (не знаю, как это сочетается с sortable; может быть и плохо)

Copy link
Collaborator Author

@Hor911 Hor911 Dec 10, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Угу, может быть плохо, надо тестить

TABLED() {str << "N/A";}
TABLED() {str << "N/A";}
TABLED() {str << "N/A";}
TABLED() {str << "N/A";}
}
}
}
}
}

str << Endl << "Output Channels:" << Endl;
TABLE_SORTABLE_CLASS("table table-condensed") {
TABLEHEAD() {
TABLER() {
TABLEH_ATTRS({{"title", "ChannelId"}}) {str << "Id";}
TABLEH_ATTRS({{"title", "DstStageId"}}) {str << "Dst";}
TABLEH() {str << "PeerId";}
TABLEH_ATTRS({{"title", "Finished"}}) {str << "F";}
TABLEH_ATTRS({{"title", "EarlyFinish"}}) {str << "EF";}
TABLEH() {str << "Push.Bytes";}
TABLEH() {str << "Push.Rows";}
TABLEH() {str << "Pop.Bytes";}
TABLEH() {str << "Pop.Rows";}
}
}
TABLEBODY() {
for (const auto& [id, info]: OutputChannelsMap) {
TABLER() {
TABLED() {str << info.ChannelId;}
TABLED() {str << info.DstStageId;}
TABLED() {
if (info.HasPeer) {
HREF(TStringBuilder() << "/node/" << info.PeerId.NodeId() << "/actors/kqp_node?ca=" << info.PeerId) {
str << info.PeerId;
}
} else {
str << "N/A";
}
}
TABLED() {str << info.Finished;}
TABLED() {str << info.EarlyFinish;}

auto channel = info.Channel;
if (!channel) {
auto stats = GetTaskRunnerStats();
if (stats) {
auto stageIt = stats->OutputChannels.find(info.DstStageId);
if (stageIt != stats->OutputChannels.end()) {
auto channelIt = stageIt->second.find(info.ChannelId);
if (channelIt != stageIt->second.end()) {
channel = channelIt->second;
}
}
}
}

if (channel) {
auto& pushStats = channel->GetPushStats();
TABLED() {str << pushStats.Bytes;}
TABLED() {str << pushStats.Rows;}
auto& popStats = channel->GetPopStats();
TABLED() {str << popStats.Bytes;}
TABLED() {str << popStats.Rows;}
} else {
TABLED() {str << "N/A";}
TABLED() {str << "N/A";}
TABLED() {str << "N/A";}
TABLED() {str << "N/A";}
}
}
}
}
}
}
}
}

void OnMonitoringPage(NActors::NMon::TEvHttpInfo::TPtr& ev) {
TStringStream str;

const TCgiParameters &cgi = ev->Get()->Request.GetParams();
auto view = cgi.Get("view");
if (view == "dump") {
DumpForMonitoring(str);
} else if (view == "run") {
if (this->Running) {
this->DoExecute();
}
DefaultMonitoringPage(str);
} else {
DefaultMonitoringPage(str);
}

this->Send(ev->Sender, new NActors::NMon::TEvHttpInfoRes(str.Str()));
}

protected:
Expand Down Expand Up @@ -1930,6 +2110,7 @@ class TDqComputeActorBase : public NActors::TActorBootstrapped<TDerived>
auto result = InputChannelsMap.emplace(
channel.GetId(),
TInputChannelInfo(
i,
LogPrefix,
channel.GetId(),
channel.GetSrcStageId(),
Expand Down Expand Up @@ -1967,7 +2148,10 @@ class TDqComputeActorBase : public NActors::TActorBootstrapped<TDerived>
for (auto& channel : outputDesc.GetChannels()) {
TOutputChannelInfo outputChannel(channel.GetId(), channel.GetDstStageId());
outputChannel.HasPeer = channel.GetDstEndpoint().HasActorId();
outputChannel.IsTransformOutput = outputDesc.HasTransform();
if (outputChannel.HasPeer) {
outputChannel.PeerId = NActors::ActorIdFromProto(channel.GetDstEndpoint().GetActorId());
}
outputChannel.IsTransformOutput = outputDesc.HasTransform();
outputChannel.WatermarksMode = channel.GetWatermarksMode();

if (Y_UNLIKELY(RuntimeSettings.StatsMode >= NDqProto::DQ_STATS_MODE_PROFILE)) {
Expand Down Expand Up @@ -2366,6 +2550,7 @@ class TDqComputeActorBase : public NActors::TActorBootstrapped<TDerived>
bool DataWasSent = false;
bool AllOutputsFinished = true;
ERunStatus LastRunStatus = ERunStatus::PendingInput;
TInstant LastRunTime;
bool LastPopReturnedNoData = false;
};
TProcessOutputsState ProcessOutputsState;
Expand Down
5 changes: 4 additions & 1 deletion ydb/library/yql/dq/actors/compute/ut/dq_async_compute_actor_ut.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,6 +45,9 @@ struct TMockHttpRequest : NMonitoring::IMonHttpRequest {
TStringStream Out;
TCgiParameters Params;
THttpHeaders Headers;
TMockHttpRequest() {
Params.Scan("view=dump");
}
IOutputStream& Output() override {
return Out;
}
Expand Down Expand Up @@ -639,8 +642,8 @@ struct TAsyncCATestFixture: public NUnitTest::TBaseFixture {
}

void DumpMonPage(auto asyncCA, auto hook) {
TMockHttpRequest request;
{
TMockHttpRequest request;
auto evHttpInfo = MakeHolder<NActors::NMon::TEvHttpInfo>(request);
ActorSystem.Send(asyncCA, EdgeActor, evHttpInfo.Release());
}
Expand Down
5 changes: 4 additions & 1 deletion ydb/library/yql/dq/actors/compute/ut/dq_sync_compute_actor_ut.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,9 @@ struct TMockHttpRequest : NMonitoring::IMonHttpRequest {
TStringStream Out;
TCgiParameters Params;
THttpHeaders Headers;
TMockHttpRequest() {
Params.Scan("view=dump");
}
IOutputStream& Output() override {
return Out;
}
Expand Down Expand Up @@ -654,8 +657,8 @@ struct TSyncComputeActorTestFixture: public NUnitTest::TBaseFixture {
}

void DumpMonPage(auto syncCA, auto hook) {
TMockHttpRequest request;
{
TMockHttpRequest request;
auto evHttpInfo = MakeHolder<NActors::NMon::TEvHttpInfo>(request);
ActorSystem.Send(syncCA, EdgeActor, evHttpInfo.Release());
}
Expand Down
Loading