Performance logging and counters:

* Tally and duration counters for Job Queue tasks and RPC calls
    optionally rendered by server_info and server_state, and
    optionally printed to a distinct log file.
    - Tally each Job Queue task as it is queued, starts, and
      finishes running. Track total duration queued and running.
    - Tally each RPC call as it starts and either finishes
      successfully or throws an exception. Track total running
      duration for each.
  * Track currently executing Job Queue tasks and RPC methods
    along with durations.
  * Json-formatted performance log file written by a dedicated
    thread, for above-described data.
  * New optional parameter, "counters", for server_info and
    server_state. If set, render Job Queue and RPC call counters
    as well as currently executing tasks.
  * New configuration section, "[perf]", to optionally control
    performance logging to a file.
  * Support optional sub-second periods when rendering human-readable
    time points.
This commit is contained in:
Mark Travis
2018-01-13 04:02:43 -08:00
committed by Nikolaos D. Bougalis
parent ef3bc92b82
commit 8eb8c77886
45 changed files with 10379 additions and 577 deletions

File diff suppressed because it is too large Load Diff

View File

@@ -40,7 +40,7 @@ class ByzantineFailureSim_test : public beast::unit_test::suite
ConsensusParms const parms{};
SimDuration const delay =
round<milliseconds>(0.2 * parms.ledgerGRANULARITY);
date::round<milliseconds>(0.2 * parms.ledgerGRANULARITY);
PeerGroup a = sim.createGroup(1);
PeerGroup b = sim.createGroup(1);
PeerGroup c = sim.createGroup(1);

View File

@@ -146,7 +146,7 @@ public:
// Connected trust and network graphs with single fixed delay
peers.trustAndConnect(
peers, round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
peers, date::round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
// everyone submits their own ID as a TX
for (Peer * p : peers)
@@ -194,10 +194,11 @@ public:
// Fast and slow network connections
fast.connect(
fast, round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
fast, date::round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
slow.connect(
network, round<milliseconds>(1.1 * parms.ledgerGRANULARITY));
network,
date::round<milliseconds>(1.1 * parms.ledgerGRANULARITY));
// All peers submit their own ID as a transaction
for (Peer* peer : network)
@@ -250,10 +251,12 @@ public:
// Fast and slow network connections
fast.connect(
fast, round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
fast,
date::round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
slow.connect(
network, round<milliseconds>(1.1 * parms.ledgerGRANULARITY));
network,
date::round<milliseconds>(1.1 * parms.ledgerGRANULARITY));
for (Peer* peer : slow)
peer->runAsValidator = isParticipant;
@@ -377,7 +380,7 @@ public:
network.trust(network);
network.connect(
network, round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
network, date::round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
// Run consensus without skew until we have a short close time
// resolution
@@ -447,7 +450,7 @@ public:
PeerGroup network = minority + majority;
SimDuration delay =
round<milliseconds>(0.2 * parms.ledgerGRANULARITY);
date::round<milliseconds>(0.2 * parms.ledgerGRANULARITY);
minority.trustAndConnect(minority + majorityA, delay);
majority.trustAndConnect(majority, delay);
@@ -552,7 +555,8 @@ public:
PeerGroup network = loner + clique;
network.connect(
network, round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
network,
date::round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
// initial round to set prior state
sim.run(1);
@@ -606,9 +610,10 @@ public:
// Fast and slow network connections
fast.connect(
fast, round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
fast, date::round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
slow.connect(
network, round<milliseconds>(1.1 * parms.ledgerGRANULARITY));
network,
date::round<milliseconds>(1.1 * parms.ledgerGRANULARITY));
// Run to the ledger *prior* to decreasing the resolution
sim.run(increaseLedgerTimeResolutionEvery - 2);
@@ -757,7 +762,7 @@ public:
PeerGroup network = a + b;
SimDuration delay =
round<milliseconds>(0.2 * parms.ledgerGRANULARITY);
date::round<milliseconds>(0.2 * parms.ledgerGRANULARITY);
a.trustAndConnect(a, delay);
b.trustAndConnect(b, delay);
@@ -804,7 +809,7 @@ public:
center.trust(validators);
SimDuration delay =
round<milliseconds>(0.2 * parms.ledgerGRANULARITY);
date::round<milliseconds>(0.2 * parms.ledgerGRANULARITY);
validators.connect(center, delay);
center[0]->runAsValidator = false;
@@ -928,8 +933,10 @@ public:
PeerGroup groupNotFastC = groupABD + groupCsplit;
PeerGroup network = groupABD + groupCsplit + groupCfast;
SimDuration delay = round<milliseconds>(0.2 * parms.ledgerGRANULARITY);
SimDuration fDelay = round<milliseconds>(0.1 * parms.ledgerGRANULARITY);
SimDuration delay = date::round<milliseconds>(
0.2 * parms.ledgerGRANULARITY);
SimDuration fDelay = date::round<milliseconds>(
0.1 * parms.ledgerGRANULARITY);
network.trust(network);
// C must have a shorter delay to see all the validations before the

View File

@@ -57,7 +57,7 @@ class ScaleFreeSim_test : public beast::unit_test::suite
// nodes with a trust line in either direction are network-connected
network.connectFromTrust(
round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
date::round<milliseconds>(0.2 * parms.ledgerGRANULARITY));
// Initialize collectors to track statistics to report
TxCollector txCollector;

View File

@@ -19,9 +19,66 @@ OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
#include <ripple/core/impl/Workers.h>
#include <ripple/beast/unit_test.h>
#include <ripple/basics/PerfLog.h>
#include <ripple/core/JobTypes.h>
#include <ripple/json/json_value.h>
#include <chrono>
#include <cstdint>
#include <memory>
#include <string>
namespace ripple {
/**
* Dummy class for unit tests.
*/
namespace perf {
class PerfLogTest
: public PerfLog
{
void rpcStart(std::string const &method, std::uint64_t requestId) override
{}
void rpcFinish(std::string const &method, std::uint64_t requestId) override
{}
void rpcError(std::string const &method, std::uint64_t dur) override
{}
void jobQueue(JobType const type) override
{}
void jobStart(JobType const type,
std::chrono::microseconds dur,
std::chrono::time_point<std::chrono::steady_clock> startTime,
int instance) override
{}
void jobFinish(JobType const type, std::chrono::microseconds dur,
int instance) override
{}
Json::Value countersJson() const override
{
return Json::Value();
}
Json::Value currentJson() const override
{
return Json::Value();
}
void resizeJobs(int const resize) override
{}
void rotate() override
{}
};
} // perf
//------------------------------------------------------------------------------
class Workers_test : public beast::unit_test::suite
@@ -35,7 +92,7 @@ public:
{
}
void processTask()
void processTask(int instance) override
{
if (--count == 0)
finished.signal();
@@ -51,8 +108,10 @@ public:
" -> " + std::to_string(tc2) + " -> " + std::to_string(tc3));
TestCallback cb;
std::unique_ptr<perf::PerfLog> perfLog =
std::make_unique<perf::PerfLogTest>();
Workers w(cb, "Test", tc1);
Workers w(cb, *perfLog, "Test", tc1);
BEAST_EXPECT(w.getNumberOfThreads() == tc1);
auto testForThreadCount = [this, &cb, &w] (int const threadCount)

View File

@@ -104,7 +104,7 @@ pretty_time(std::ostream& os, std::chrono::duration<Rep, Period> d)
else
{
// use integral
os << round<nanoseconds>(d).count();
os << date::round<nanoseconds>(d).count();
}
os << "ns";
}
@@ -120,7 +120,7 @@ pretty_time(std::ostream& os, std::chrono::duration<Rep, Period> d)
else
{
// use integral
os << round<microseconds>(d).count();
os << date::round<microseconds>(d).count();
}
os << "us";
}
@@ -136,7 +136,7 @@ pretty_time(std::ostream& os, std::chrono::duration<Rep, Period> d)
else
{
// use integral
os << round<milliseconds>(d).count();
os << date::round<milliseconds>(d).count();
}
os << "ms";
}
@@ -152,7 +152,7 @@ pretty_time(std::ostream& os, std::chrono::duration<Rep, Period> d)
else
{
// use integral
os << round<seconds>(d).count();
os << date::round<seconds>(d).count();
}
os << "s";
}
@@ -168,7 +168,7 @@ pretty_time(std::ostream& os, std::chrono::duration<Rep, Period> d)
else
{
// use integral
os << round<minutes>(d).count();
os << date::round<minutes>(d).count();
}
os << "min";
}

View File

@@ -26,6 +26,7 @@
#include <test/basics/hardened_hash_test.cpp>
#include <test/basics/KeyCache_test.cpp>
#include <test/basics/mulDiv_test.cpp>
#include <test/basics/PerfLog_test.cpp>
#include <test/basics/RangeSet_test.cpp>
#include <test/basics/Slice_test.cpp>
#include <test/basics/StringUtilities_test.cpp>