Merge branch 'release-6.3'

# Conflicts:
#	fdbserver/fdbserver.actor.cpp
This commit is contained in:
Steve Atherton 2021-02-15 02:13:06 -08:00
commit 76a4293d1d
7 changed files with 34 additions and 17 deletions

View File

@ -2,6 +2,10 @@
Release Notes
#############
6.2.31
======
* Fix a rare invalid memory access on data distributor when snapshotting large clusters. This is a follow up to `PR #4076 <https://github.com/apple/foundationdb/pull/4076>`_. `(PR #4317) <https://github.com/apple/foundationdb/pull/4317>`_
6.2.30
======
* A storage server which has fallen behind will deprioritize reads in order to catch up. This change causes some saturating workloads to experience high read latencies instead of high GRV latencies. `(PR #4218) <https://github.com/apple/foundationdb/pull/4218>`_

View File

@ -2234,7 +2234,7 @@ TEST_CASE("/ManagementAPI/AutoQuorumChange/checkLocality") {
data.address.ip = IPAddress(i);
if(g_network->isSimulated()) {
g_simulator.newProcess(format("TestProcess%d", i).c_str(), data.address.ip, data.address.port, false, 1,
g_simulator.newProcess("TestCoordinator", data.address.ip, data.address.port, false, 1,
data.locality, ProcessClass(ProcessClass::CoordinatorClass, ProcessClass::CommandLineSource),
"", "", currentProtocolVersion);
}

View File

@ -241,7 +241,7 @@ ACTOR Future<Void> dataDistributionTracker(Reference<InitialDataDistribution> in
FutureStream<Promise<int64_t>> getAverageShardBytes,
Promise<Void> readyToStart, Reference<AsyncVar<bool>> zeroHealthyTeams,
UID distributorId, KeyRangeMap<ShardTrackedData>* shards,
bool const* trackerCancelled);
bool* trackerCancelled);
ACTOR Future<Void> dataDistributionQueue(
Database cx,

View File

@ -94,7 +94,7 @@ struct DataDistributionTracker {
// The reference to trackerCancelled must be extracted by actors,
// because by the time (trackerCancelled == true) this memory cannot
// be accessed
bool const& trackerCancelled;
bool& trackerCancelled;
// This class extracts the trackerCancelled reference from a DataDistributionTracker object
// Because some actors spawned by the dataDistributionTracker outlive the DataDistributionTracker
@ -123,7 +123,7 @@ struct DataDistributionTracker {
PromiseStream<RelocateShard> const& output,
Reference<ShardsAffectedByTeamFailure> shardsAffectedByTeamFailure,
Reference<AsyncVar<bool>> anyZeroHealthyTeams, KeyRangeMap<ShardTrackedData>& shards,
bool const& trackerCancelled)
bool& trackerCancelled)
: cx(cx), distributorId(distributorId), dbSizeEstimate(new AsyncVar<int64_t>()), systemSizeEstimate(0),
maxShardSize(new AsyncVar<Optional<int64_t>>()), sizeChanges(false), readyToStart(readyToStart), output(output),
shardsAffectedByTeamFailure(shardsAffectedByTeamFailure), anyZeroHealthyTeams(anyZeroHealthyTeams),
@ -131,6 +131,7 @@ struct DataDistributionTracker {
~DataDistributionTracker()
{
trackerCancelled = true;
//Cancel all actors so they aren't waiting on sizeChanged broken promise
sizeChanges.clear(false);
}
@ -903,7 +904,7 @@ ACTOR Future<Void> dataDistributionTracker(Reference<InitialDataDistribution> in
FutureStream<Promise<int64_t>> getAverageShardBytes,
Promise<Void> readyToStart, Reference<AsyncVar<bool>> anyZeroHealthyTeams,
UID distributorId, KeyRangeMap<ShardTrackedData>* shards,
bool const* trackerCancelled) {
bool* trackerCancelled) {
state DataDistributionTracker self(cx, distributorId, readyToStart, output, shardsAffectedByTeamFailure,
anyZeroHealthyTeams, *shards, *trackerCancelled);
state Future<Void> loggingTrigger = Void();

View File

@ -86,13 +86,25 @@ Error internal_error_impl(const char* a_nm, long long a, const char * op_nm, con
return Error(error_code_internal_error);
}
Error::Error(int error_code)
: error_code(error_code), flags(0)
{
Error::Error(int error_code) : error_code(error_code), flags(0) {
if (TRACE_SAMPLE()) TraceEvent(SevSample, "ErrorCreated").detail("ErrorCode", error_code);
//std::cout << "Error: " << error_code << std::endl;
// std::cout << "Error: " << error_code << std::endl;
if (error_code >= 3000 && error_code < 6000) {
TraceEvent(SevError, "SystemError").error(*this).backtrace();
{
TraceEvent te(SevError, "SystemError");
te.error(*this).backtrace();
if (error_code == error_code_unknown_error) {
auto exception = std::current_exception();
if (exception) {
try {
std::rethrow_exception(exception);
} catch (std::exception& e) {
te.detail("StdException", e.what());
} catch (...) {
}
}
}
}
if (g_crashOnError) {
flushOutputStreams();
flushTraceFileVoid();

View File

@ -778,7 +778,7 @@ bool TLSPolicy::verify_peer(bool preverified, X509_STORE_CTX* store_ctx) {
}
if(!preverified) {
TraceEvent("TLSPolicyFailure").suppressFor(1.0).detail("Reason", "preverification failed").detail("VerifyError", X509_verify_cert_error_string(X509_STORE_CTX_get_error(store_ctx)));
TraceEvent(SevWarn, "TLSPolicyFailure").suppressFor(1.0).detail("Reason", "preverification failed").detail("VerifyError", X509_verify_cert_error_string(X509_STORE_CTX_get_error(store_ctx)));
return false;
}
@ -801,7 +801,7 @@ bool TLSPolicy::verify_peer(bool preverified, X509_STORE_CTX* store_ctx) {
if (!rc) {
// log the various failure reasons
for (std::string reason : verify_failure_reasons) {
TraceEvent("TLSPolicyFailure").suppressFor(1.0).detail("Reason", reason);
TraceEvent(SevWarn, "TLSPolicyFailure").suppressFor(1.0).detail("Reason", reason);
}
}
return rc;

View File

@ -859,7 +859,7 @@ bool TraceEvent::init() {
detail("Severity", int(severity));
detail("Time", "0.000000");
timeIndex = fields.size() - 1;
if (FLOW_KNOBS->TRACE_DATETIME_ENABLED) {
if (FLOW_KNOBS && FLOW_KNOBS->TRACE_DATETIME_ENABLED) {
detail("DateTime", "");
}
@ -1081,7 +1081,7 @@ void TraceEvent::log() {
if (enabled) {
double time = TraceEvent::getCurrentTime();
fields.mutate(timeIndex).second = format("%.6f", time);
if (FLOW_KNOBS->TRACE_DATETIME_ENABLED) {
if (FLOW_KNOBS && FLOW_KNOBS->TRACE_DATETIME_ENABLED) {
fields.mutate(timeIndex + 1).second = TraceEvent::printRealTime(time);
}
@ -1265,7 +1265,7 @@ void TraceBatch::dump() {
TraceBatch::EventInfo::EventInfo(double time, const char *name, uint64_t id, const char *location) {
fields.addField("Severity", format("%d", (int)TRACE_BATCH_IMPLICIT_SEVERITY));
fields.addField("Time", format("%.6f", time));
if (FLOW_KNOBS->TRACE_DATETIME_ENABLED) {
if (FLOW_KNOBS && FLOW_KNOBS->TRACE_DATETIME_ENABLED) {
fields.addField("DateTime", TraceEvent::printRealTime(time));
}
fields.addField("Type", name);
@ -1276,7 +1276,7 @@ TraceBatch::EventInfo::EventInfo(double time, const char *name, uint64_t id, con
TraceBatch::AttachInfo::AttachInfo(double time, const char *name, uint64_t id, uint64_t to) {
fields.addField("Severity", format("%d", (int)TRACE_BATCH_IMPLICIT_SEVERITY));
fields.addField("Time", format("%.6f", time));
if (FLOW_KNOBS->TRACE_DATETIME_ENABLED) {
if (FLOW_KNOBS && FLOW_KNOBS->TRACE_DATETIME_ENABLED) {
fields.addField("DateTime", TraceEvent::printRealTime(time));
}
fields.addField("Type", name);
@ -1287,7 +1287,7 @@ TraceBatch::AttachInfo::AttachInfo(double time, const char *name, uint64_t id, u
TraceBatch::BuggifyInfo::BuggifyInfo(double time, int activated, int line, std::string file) {
fields.addField("Severity", format("%d", (int)TRACE_BATCH_IMPLICIT_SEVERITY));
fields.addField("Time", format("%.6f", time));
if (FLOW_KNOBS->TRACE_DATETIME_ENABLED) {
if (FLOW_KNOBS && FLOW_KNOBS->TRACE_DATETIME_ENABLED) {
fields.addField("DateTime", TraceEvent::printRealTime(time));
}
fields.addField("Type", "BuggifySection");