feat(telemetry): add preflight/preclaim stage spans and stage attribute

The tx.transactor span covered only the apply stage; preflight and
preclaim had no telemetry, so a transaction that hard-failed those
stages produced no apply-pipeline span and per-stage latency/failure
was invisible.

Add tx.preflight and tx.preclaim spans in applySteps.cpp via a
makeStageSpan() helper using SpanGuard::hashSpan, so all three stages
share a deterministic trace_id derived from txID[0:16] even though they
run sequentially and often cross-thread. Each span carries stage,
tx_type, and ter_result; exceptions are recorded as tefEXCEPTION before
the public wrappers map them. The type lookup is guarded behind the
span-active check so it costs nothing when tracing is off.

Add a stage="apply" attribute to the tx.transactor span and move its
three hardcoded attribute strings to a new library-safe header
include/xrpl/tx/detail/TxApplySpanNames.h, which mirrors the daemon-side
TxSpanNames.h strings so the collector spanmetrics connector aggregates
both span sets under one dimension set.

A constants-contract test pins the span-name, attribute-key, and
stage-value strings; span content stays covered by the docker
integration test, as the rest of the telemetry suite is.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
Pratik Mankawde
2026-06-05 11:11:55 +01:00
parent dfd67b8124
commit 6428c9f13c
5 changed files with 282 additions and 37 deletions

View File

@@ -240,6 +240,7 @@ xrpl.tx > xrpl.basics
xrpl.tx > xrpl.core
xrpl.tx > xrpl.ledger
xrpl.tx > xrpl.protocol
xrpl.tx > xrpl.telemetry
xrpld.app > test.unit_test
xrpld.app > xrpl.basics
xrpld.app > xrpl.core

View File

@@ -0,0 +1,109 @@
#pragma once
/** Compile-time span name constants for the transaction apply pipeline.
*
* Defines the span names and attribute keys used by the three apply-pipeline
* stages — preflight, preclaim, and transactor (apply) — that run inside the
* library (`src/libxrpl/tx/`). Built on the StaticStr/join() primitives from
* <xrpl/telemetry/SpanNames.h>.
*
* Why a separate header from TxSpanNames.h:
* TxSpanNames.h lives under src/xrpld/ (daemon) and serves the overlay/app
* lifecycle spans (tx.receive, tx.process). Library code (applySteps.cpp,
* Transactor.cpp) must not depend on daemon headers, so the apply-pipeline
* constants live here instead. The attribute strings ("tx_type",
* "ter_result", "applied") intentionally match TxSpanNames.h so the collector
* spanmetrics connector aggregates both sets under the same dimensions.
*
* Span hierarchy (deterministic trace_id derived from txID[0:16]):
*
* The three stages run sequentially and often on different threads, so they
* do not auto-parent. Each uses a hash-derived trace_id keyed on the same
* transaction id, placing all three under one trace without context
* propagation. A transaction that hard-fails preflight or preclaim never
* reaches the transactor span — the stage attribute identifies where it
* stopped.
*
* +-----------------------------------------------------------+
* | trace_id = txID[0:16] |
* | |
* | +-------------------+ +------------------+ +-------+ |
* | | tx.preflight | | tx.preclaim | | tx. | |
* | | stage=preflight |-->| stage=preclaim |-->| trans | |
* | | tx_type | | tx_type | | actor | |
* | | ter_result | | ter_result | | stage=| |
* | +-------------------+ +------------------+ | apply | |
* | stateless checks ledger-aware checks +-------+ |
* | (signature, fields) (sequence, fee) applies |
* +-----------------------------------------------------------+
*
* Usage:
* @code
* #include <xrpl/tx/detail/TxApplySpanNames.h>
* using namespace telemetry;
*
* // preflight() / preclaim() use hashSpan with a full span name:
* auto span = SpanGuard::hashSpan(
* TraceCategory::Transactions, tx_apply_span::preflight,
* txID.data(), txID.kBytes);
* span.setAttribute(tx_apply_span::attr::stage, tx_apply_span::val::preflight);
* span.setAttribute(tx_apply_span::attr::terResult, transToken(ter).c_str());
* @endcode
*
* @code
* // Transactor::operator() uses span() with prefix + suffix:
* auto span = SpanGuard::span(
* TraceCategory::Transactions, seg::tx, tx_apply_span::op::transactor);
* span.setAttribute(tx_apply_span::attr::stage, tx_apply_span::val::apply);
* @endcode
*/
#include <xrpl/telemetry/SpanNames.h>
namespace xrpl::telemetry::tx_apply_span {
// ===== Span operation suffixes =============================================
namespace op {
/// "preflight" — stateless transaction checks (suffix form).
inline constexpr auto preflight = makeStr("preflight");
/// "preclaim" — ledger-aware checks before fee claim (suffix form).
inline constexpr auto preclaim = makeStr("preclaim");
/// "transactor" — the apply stage (suffix form, used with span()).
inline constexpr auto transactor = makeStr("transactor");
} // namespace op
// ===== Full span names (tx.<op>) ===========================================
/// "tx.preflight" — full name for hashSpan() at the preflight stage.
inline constexpr auto preflight = join(seg::tx, op::preflight);
/// "tx.preclaim" — full name for hashSpan() at the preclaim stage.
inline constexpr auto preclaim = join(seg::tx, op::preclaim);
// ===== Attribute keys ======================================================
namespace attr {
/// "stage" — which apply-pipeline stage this span represents. Drives the
/// collector spanmetrics `stage` dimension for per-stage RED metrics.
inline constexpr auto stage = makeStr("stage");
/// "tx_type" — transaction type name (e.g., "Payment", "OfferCreate").
/// Matches tx_span::attr::txType so both share the spanmetrics dimension.
inline constexpr auto txType = makeStr("tx_type");
/// "ter_result" — engine result code after the stage (e.g., "tesSUCCESS").
inline constexpr auto terResult = makeStr("ter_result");
/// "applied" — whether the transaction was applied to the ledger (apply only).
inline constexpr auto applied = makeStr("applied");
} // namespace attr
// ===== Attribute values (stage names) ======================================
namespace val {
/// "preflight" — value of the stage attribute on tx.preflight.
inline constexpr auto preflight = makeStr("preflight");
/// "preclaim" — value of the stage attribute on tx.preclaim.
inline constexpr auto preclaim = makeStr("preclaim");
/// "apply" — value of the stage attribute on tx.transactor.
inline constexpr auto apply = makeStr("apply");
} // namespace val
} // namespace xrpl::telemetry::tx_apply_span

View File

@@ -44,6 +44,7 @@
#include <xrpl/tx/SignerEntries.h>
#include <xrpl/tx/apply.h>
#include <xrpl/tx/applySteps.h>
#include <xrpl/tx/detail/TxApplySpanNames.h>
#include <cstddef>
#include <cstdint>
@@ -1199,9 +1200,11 @@ Transactor::operator()()
auto span = telemetry::SpanGuard::span(
telemetry::TraceCategory::Transactions,
telemetry::seg::tx,
telemetry::makeStr("transactor"));
telemetry::tx_apply_span::op::transactor);
// "apply" — the third apply-pipeline stage, after preflight and preclaim.
span.setAttribute(telemetry::tx_apply_span::attr::stage, telemetry::tx_apply_span::val::apply);
if (auto const* fmt = TxFormats::getInstance().findByType(ctx_.tx.getTxnType()))
span.setAttribute("tx_type", fmt->getName().c_str());
span.setAttribute(telemetry::tx_apply_span::attr::txType, fmt->getName().c_str());
JLOG(j_.trace()) << "apply: " << ctx_.tx.getTransactionID();
@@ -1429,8 +1432,8 @@ Transactor::operator()()
JLOG(j_.trace()) << (applied ? "applied " : "not applied ") << transToken(result);
span.setAttribute("ter_result", transToken(result).c_str());
span.setAttribute("applied", applied);
span.setAttribute(telemetry::tx_apply_span::attr::terResult, transToken(result).c_str());
span.setAttribute(telemetry::tx_apply_span::attr::applied, applied);
return {result, applied, metadata};
}

View File

@@ -13,13 +13,16 @@
#include <xrpl/protocol/SeqProxy.h>
#include <xrpl/protocol/TER.h>
#include <xrpl/protocol/XRPAmount.h>
#include <xrpl/telemetry/SpanGuard.h>
#include <xrpl/tx/ApplyContext.h>
#include <xrpl/tx/Transactor.h>
#include <xrpl/tx/detail/TxApplySpanNames.h>
#include <cstdint>
#include <exception>
#include <memory>
#include <optional>
#include <string_view>
#include <utility>
#pragma push_macro("TRANSACTION")
#undef TRANSACTION
@@ -51,6 +54,47 @@ struct UnknownTxnType : std::exception
}
};
/** Look up the human-readable transaction type name for span attributes.
* Returns nullptr if the type is unknown so the caller can skip the
* attribute rather than emit an empty value.
*/
char const*
txTypeName(TxType txnType)
{
if (auto const* fmt = TxFormats::getInstance().findByType(txnType))
return fmt->getName().c_str();
return nullptr;
}
/** Create a deterministic-trace span for an apply-pipeline stage.
*
* The trace_id is derived from txID[0:16] so the preflight, preclaim, and
* transactor spans of one transaction share a trace even though they run
* sequentially and often on different threads. Sets the stage, tx_type, and
* (after the stage runs) ter_result attributes that drive the collector
* spanmetrics dimensions. A no-op when telemetry is disabled.
*
* @param name Full span name (tx_apply_span::preflight / ::preclaim).
* @param stage Stage attribute value (tx_apply_span::val::*).
* @param tx The transaction supplying the id and type.
*/
[[nodiscard]] telemetry::SpanGuard
makeStageSpan(std::string_view name, std::string_view stage, STTx const& tx)
{
auto const txID = tx.getTransactionID();
auto span = telemetry::SpanGuard::hashSpan(
telemetry::TraceCategory::Transactions, name, txID.data(), txID.kBytes);
// Guard the type lookup behind the active check: preflight runs for every
// transaction, so findByType() must not run when tracing is off/disabled.
if (span)
{
span.setAttribute(telemetry::tx_apply_span::attr::stage, stage);
if (char const* typeName = txTypeName(tx.getTxnType()))
span.setAttribute(telemetry::tx_apply_span::attr::txType, typeName);
}
return span;
}
// Call a lambda with the concrete transaction type as a template parameter
// throw an "UnknownTxnType" exception on error
template <class F>
@@ -133,82 +177,118 @@ consequencesHelper(PreflightContext const& ctx)
static std::pair<NotTEC, TxConsequences>
invokePreflight(PreflightContext const& ctx)
{
// Trace the preflight stage. The span shares the transaction's
// deterministic trace_id so it correlates with preclaim and transactor.
auto span = makeStageSpan(
telemetry::tx_apply_span::preflight, telemetry::tx_apply_span::val::preflight, ctx.tx);
try
{
return withTxnType(ctx.rules, ctx.tx.getTxnType(), [&]<typename T>() {
auto result = withTxnType(ctx.rules, ctx.tx.getTxnType(), [&]<typename T>() {
auto const tec = Transactor::invokePreflight<T>(ctx);
return std::make_pair(
tec, isTesSuccess(tec) ? consequencesHelper<T>(ctx) : TxConsequences{tec});
});
if (span)
span.setAttribute(
telemetry::tx_apply_span::attr::terResult, transToken(result.first).c_str());
return result;
}
catch (UnknownTxnType const& e)
{
// Should never happen
// LCOV_EXCL_START
JLOG(ctx.j.fatal()) << "Unknown transaction type in preflight: " << e.txnType;
span.recordException(e);
UNREACHABLE("xrpl::invokePreflight : unknown transaction type");
return {temUNKNOWN, TxConsequences{temUNKNOWN}};
// LCOV_EXCL_STOP
}
catch (std::exception const& e)
{
// The caller's preflight() maps this to tefEXCEPTION. Record it on the
// span before unwinding so per-stage error counts include exceptions.
span.setAttribute(
telemetry::tx_apply_span::attr::terResult, transToken(tefEXCEPTION).c_str());
span.recordException(e);
throw;
}
}
static TER
invokePreclaim(PreclaimContext const& ctx)
{
// Trace the preclaim stage under the transaction's deterministic trace_id.
auto span = makeStageSpan(
telemetry::tx_apply_span::preclaim, telemetry::tx_apply_span::val::preclaim, ctx.tx);
try
{
// use name hiding to accomplish compile-time polymorphism of static
// class functions for Transactor and derived classes.
return withTxnType(ctx.view.rules(), ctx.tx.getTxnType(), [&]<typename T>() -> TER {
// preclaim functionality is divided into two sections:
// 1. Up to and including the signature check: returns NotTEC.
// All transaction checks before and including checkSign
// MUST return NotTEC, or something more restrictive.
// Allowing tec results in these steps risks theft or
// destruction of funds, as a fee will be charged before the
// signature is checked.
// 2. After the signature check: returns TER.
TER const preclaimTer =
withTxnType(ctx.view.rules(), ctx.tx.getTxnType(), [&]<typename T>() -> TER {
// preclaim functionality is divided into two sections:
// 1. Up to and including the signature check: returns NotTEC.
// All transaction checks before and including checkSign
// MUST return NotTEC, or something more restrictive.
// Allowing tec results in these steps risks theft or
// destruction of funds, as a fee will be charged before the
// signature is checked.
// 2. After the signature check: returns TER.
// If the transactor requires a valid account and the
// transaction doesn't list one, preflight will have already
// a flagged a failure.
auto const id = ctx.tx.getAccountID(sfAccount);
// If the transactor requires a valid account and the
// transaction doesn't list one, preflight will have already
// a flagged a failure.
auto const id = ctx.tx.getAccountID(sfAccount);
if (id != beast::kZero)
{
if (NotTEC const preSigResult = [&]() -> NotTEC {
if (NotTEC const result = T::checkSeqProxy(ctx.view, ctx.tx, ctx.j))
return result;
if (id != beast::kZero)
{
if (NotTEC const preSigResult = [&]() -> NotTEC {
if (NotTEC const result = T::checkSeqProxy(ctx.view, ctx.tx, ctx.j))
return result;
if (NotTEC const result = T::checkPriorTxAndLastLedger(ctx))
return result;
if (NotTEC const result = T::checkPriorTxAndLastLedger(ctx))
return result;
if (NotTEC const result = T::checkPermission(ctx.view, ctx.tx))
return result;
if (NotTEC const result = T::checkPermission(ctx.view, ctx.tx))
return result;
if (NotTEC const result = T::checkSign(ctx))
return result;
if (NotTEC const result = T::checkSign(ctx))
return result;
return tesSUCCESS;
}())
return preSigResult;
return tesSUCCESS;
}())
return preSigResult;
if (TER const result = T::checkFee(ctx, calculateBaseFee(ctx.view, ctx.tx)))
return result;
}
if (TER const result = T::checkFee(ctx, calculateBaseFee(ctx.view, ctx.tx)))
return result;
}
return T::preclaim(ctx);
});
return T::preclaim(ctx);
});
if (span)
span.setAttribute(
telemetry::tx_apply_span::attr::terResult, transToken(preclaimTer).c_str());
return preclaimTer;
}
catch (UnknownTxnType const& e)
{
// Should never happen
// LCOV_EXCL_START
JLOG(ctx.j.fatal()) << "Unknown transaction type in preclaim: " << e.txnType;
span.recordException(e);
UNREACHABLE("xrpl::invokePreclaim : unknown transaction type");
return temUNKNOWN;
// LCOV_EXCL_STOP
}
catch (std::exception const& e)
{
// The caller's preclaim() maps this to tefEXCEPTION. Record it on the
// span before unwinding so per-stage error counts include exceptions.
span.setAttribute(
telemetry::tx_apply_span::attr::terResult, transToken(tefEXCEPTION).c_str());
span.recordException(e);
throw;
}
}
/**

View File

@@ -0,0 +1,52 @@
#include <xrpl/tx/detail/TxApplySpanNames.h>
#include <gtest/gtest.h>
#include <string_view>
/** Contract tests for the transaction apply-pipeline span constants.
*
* The span names and attribute keys in TxApplySpanNames.h are a cross-component
* contract: the collector spanmetrics connector aggregates on these exact
* strings (dimensions tx_type, ter_result, stage) and the Grafana
* transaction-overview dashboard queries them. A silent rename here would
* break per-stage metrics with no compile error, so these tests pin the
* literal values. They need no telemetry runtime and run in every build.
*/
using namespace xrpl::telemetry;
TEST(TxApplySpanNames, span_names_are_dot_qualified)
{
// Full span names feed SpanGuard::hashSpan() in applySteps.cpp.
EXPECT_EQ(std::string_view(tx_apply_span::preflight), "tx.preflight");
EXPECT_EQ(std::string_view(tx_apply_span::preclaim), "tx.preclaim");
}
TEST(TxApplySpanNames, operation_suffixes)
{
// Suffix used with SpanGuard::span(cat, seg::tx, suffix) in Transactor.cpp.
EXPECT_EQ(std::string_view(tx_apply_span::op::preflight), "preflight");
EXPECT_EQ(std::string_view(tx_apply_span::op::preclaim), "preclaim");
EXPECT_EQ(std::string_view(tx_apply_span::op::transactor), "transactor");
}
TEST(TxApplySpanNames, attribute_keys_match_collector_dimensions)
{
// These keys MUST match docker/telemetry/otel-collector-config.yaml
// spanmetrics dimensions and TxSpanNames.h (so both span sets aggregate
// under one dimension).
EXPECT_EQ(std::string_view(tx_apply_span::attr::stage), "stage");
EXPECT_EQ(std::string_view(tx_apply_span::attr::txType), "tx_type");
EXPECT_EQ(std::string_view(tx_apply_span::attr::terResult), "ter_result");
EXPECT_EQ(std::string_view(tx_apply_span::attr::applied), "applied");
}
TEST(TxApplySpanNames, stage_values_are_the_three_pipeline_stages)
{
// The stage attribute carries exactly these three values; they become the
// spanmetrics `stage` dimension cardinality (3) and the dashboard filter.
EXPECT_EQ(std::string_view(tx_apply_span::val::preflight), "preflight");
EXPECT_EQ(std::string_view(tx_apply_span::val::preclaim), "preclaim");
EXPECT_EQ(std::string_view(tx_apply_span::val::apply), "apply");
}