1
0
mirror of https://github.com/mariadb-corporation/mariadb-columnstore-engine.git synced 2025-04-18 21:44:02 +03:00
mariadb-columnstore-engine/versioning/BRM/sessionmanagerserver.cpp
drrtuy 444cf4c65e fix(aggregation, disk-based) MCOL-5691 distinct aggregate disk based (#3145)
* fix(aggregation, disk-based): MCOL-5689 this fixes disk-based distinct aggregation functions
Previously disk-based distinct aggregation functions produced incorrect results b/c there was no finalization applied for previous generations stored on disk.

*  fix(aggregation, disk-based): Fix disk-based COUNT(DISTINCT ...) queries. (Case 2). (Distinct & Multi-Distinct, Single- & Multi-Threaded).

* fix(aggregation, disk-based): Fix disk-based DISTINCT & GROUP BY queries. (Case 1). (Distinct & Multi-Distinct, Single- & Multi-Threaded).

---------

Co-authored-by: Theresa Hradilak <theresa.hradilak@gmail.com>
Co-authored-by: Roman Nozdrin <rnozdrin@mariadb.com>
2024-03-26 00:39:53 +03:00

461 lines
11 KiB
C++

/* Copyright (C) 2014 InfiniDB, Inc.
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; version 2 of
the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,
MA 02110-1301, USA. */
/*****************************************************************************
* $Id: sessionmanagerserver.cpp 1906 2013-06-14 19:15:32Z rdempsey $
*
****************************************************************************/
/*
* This class issues Transaction ID and keeps track of the current version ID
*/
#include <sys/types.h>
#include <sys/stat.h>
#include <cerrno>
#include <fcntl.h>
#include <mutex>
#include <unistd.h>
#include <iostream>
#include <string>
#include <stdexcept>
#include <limits>
#include <unordered_set>
using namespace std;
#include <boost/thread/mutex.hpp>
#include <boost/scoped_ptr.hpp>
using namespace boost;
#include "brmtypes.h"
#include "calpontsystemcatalog.h"
using namespace execplan;
#include "configcpp.h"
#include "atomicops.h"
#define SESSIONMANAGERSERVER_DLLEXPORT
#include "sessionmanagerserver.h"
#undef SESSIONMANAGERSERVER_DLLEXPORT
#ifndef O_BINARY
#define O_BINARY 0
#endif
#ifndef O_DIRECT
#define O_DIRECT 0
#endif
#ifndef O_LARGEFILE
#define O_LARGEFILE 0
#endif
#ifndef O_NOATIME
#define O_NOATIME 0
#endif
#include "IDBDataFile.h"
#include "IDBPolicy.h"
using namespace idbdatafile;
namespace BRM
{
const uint32_t SessionManagerServer::SS_READY = 1 << 0; // Set by dmlProc one time when dmlProc is ready
const uint32_t SessionManagerServer::SS_SUSPENDED =
1 << 1; // Set by console when the system has been suspended by user.
const uint32_t SessionManagerServer::SS_SUSPEND_PENDING =
1 << 2; // Set by console when user wants to suspend, but writing is occuring.
const uint32_t SessionManagerServer::SS_SHUTDOWN_PENDING =
1 << 3; // Set by console when user wants to shutdown, but writing is occuring.
const uint32_t SessionManagerServer::SS_ROLLBACK =
1 << 4; // In combination with a PENDING flag, force a rollback as soom as possible.
const uint32_t SessionManagerServer::SS_FORCE =
1 << 5; // In combination with a PENDING flag, force a shutdown without rollback.
const uint32_t SessionManagerServer::SS_QUERY_READY =
1 << 6; // Set by ProcManager when system is ready for queries
SessionManagerServer::SessionManagerServer() : unique32(0), unique64(0)
{
config::Config* conf;
string stmp;
const char* ctmp;
conf = config::Config::makeConfig();
try
{
stmp = conf->getConfig("SessionManager", "MaxConcurrentTransactions");
}
catch (const std::exception& e)
{
cout << e.what() << endl;
stmp.clear();
}
if (stmp != "")
{
int64_t tmp;
ctmp = stmp.c_str();
tmp = config::Config::fromText(ctmp);
if (tmp < 1)
maxTxns = 1;
else
maxTxns = static_cast<int>(tmp);
}
else
maxTxns = 1;
txnidFilename = conf->getConfig("SessionManager", "TxnIDFile");
semValue = maxTxns;
_verID = 0;
_sysCatVerID = 0;
systemState = 0;
try
{
loadState();
}
catch (...)
{
// first-time run most likely, ignore the error
}
}
SessionManagerServer::~SessionManagerServer()
{
}
void SessionManagerServer::reset()
{
mutex.try_lock();
semValue = maxTxns;
condvar.notify_all();
activeTxns.clear();
mutex.unlock();
}
void SessionManagerServer::loadState()
{
int lastTxnID;
int err;
int lastSysCatVerId;
again:
// There are now 3 pieces of info stored in the txnidfd file: last
// transaction id, last system catalog version id, and the
// system state flags. All these values are stored in shared, an
// instance of struct Overlay.
// If we fail to read a full four bytes for any value, then the
// value isn't in the file, and we start with the default.
if (IDBPolicy::exists(txnidFilename.c_str()))
{
scoped_ptr<IDBDataFile> txnidfp(IDBDataFile::open(
IDBPolicy::getType(txnidFilename.c_str(), IDBPolicy::WRITEENG), txnidFilename.c_str(), "rb", 0));
if (!txnidfp)
{
perror("SessionManagerServer(): open");
throw runtime_error("SessionManagerServer: Could not open the transaction ID file");
}
// Last transaction id
txnidfp->seek(0, SEEK_SET);
err = txnidfp->read(&lastTxnID, 4);
if (err < 0 && errno != EINTR)
{
perror("Sessionmanager::initSegment(): read");
throw runtime_error("SessionManagerServer: read failed, aborting");
}
else if (err < 0)
goto again;
else if (err == sizeof(int))
_verID = lastTxnID;
// last system catalog version id
err = txnidfp->read(&lastSysCatVerId, 4);
if (err < 0 && errno != EINTR)
{
perror("Sessionmanager::initSegment(): read");
throw runtime_error("SessionManagerServer: read failed, aborting");
}
else if (err < 0)
goto again;
else if (err == sizeof(int))
_sysCatVerID = lastSysCatVerId;
// System state. Contains flags regarding the suspend state of the system.
err = txnidfp->read(&systemState, 4);
if (err < 0 && errno == EINTR)
{
goto again;
}
else if (err == sizeof(int))
{
// Turn off the pending and force flags. They make no sense for a clean start.
// Turn off the ready flag. DMLProc will set it back on when
// initialized.
systemState &=
~(SS_READY | SS_QUERY_READY | SS_SUSPEND_PENDING | SS_SHUTDOWN_PENDING | SS_ROLLBACK | SS_FORCE);
}
else
{
// else no problem. System state wasn't saved. Might be an upgraded system.
systemState = 0;
}
}
}
/* Save the systemState flags of the Overlay
* segment. This is saved in the third
* word of txnid File
*/
void SessionManagerServer::saveSystemState()
{
saveSMTxnIDAndState();
}
const QueryContext SessionManagerServer::verID()
{
QueryContext ret;
boost::mutex::scoped_lock lk(mutex);
ret.currentScn = _verID;
for (iterator i = activeTxns.begin(); i != activeTxns.end(); ++i)
ret.currentTxns->push_back(i->second);
return ret;
}
const QueryContext SessionManagerServer::sysCatVerID()
{
QueryContext ret;
boost::mutex::scoped_lock lk(mutex);
ret.currentScn = _sysCatVerID;
for (iterator i = activeTxns.begin(); i != activeTxns.end(); ++i)
ret.currentTxns->push_back(i->second);
return ret;
}
uint32_t SessionManagerServer::newCpimportJob()
{
std::scoped_lock lk(cpimportMutex);
activeCpimportJobs.insert(cpimportJobId);
auto ret = cpimportJobId;
++cpimportJobId;
return ret;
}
void SessionManagerServer::finishCpimortJob(uint32_t jobId)
{
std::scoped_lock lk(cpimportMutex);
if (activeCpimportJobs.count(jobId))
activeCpimportJobs.erase(jobId);
}
void SessionManagerServer::clearAllCpimportJobs()
{
std::scoped_lock lk(cpimportMutex);
activeCpimportJobs.clear();
}
const TxnID SessionManagerServer::newTxnID(const SID session, bool block, bool isDDL)
{
TxnID ret; // ctor must set valid = false
iterator it;
boost::mutex::scoped_lock lk(mutex);
// if it already has a txn...
it = activeTxns.find(session);
if (it != activeTxns.end())
{
ret.id = it->second;
ret.valid = true;
return ret;
}
if (!block && semValue == 0)
return ret;
else
while (semValue == 0)
condvar.wait(lk);
semValue--;
idbassert(semValue <= (uint32_t)maxTxns);
ret.id = ++_verID;
ret.valid = true;
activeTxns[session] = ret.id;
if (isDDL)
++_sysCatVerID;
saveSMTxnIDAndState();
return ret;
}
void SessionManagerServer::finishTransaction(TxnID& txn)
{
iterator it;
boost::mutex::scoped_lock lk(mutex);
bool found = false;
if (!txn.valid)
throw invalid_argument("SessionManagerServer::finishTransaction(): transaction is invalid");
for (it = activeTxns.begin(); it != activeTxns.end();)
{
if (it->second == txn.id)
{
activeTxns.erase(it++);
txn.valid = false;
found = true;
// we could probably break at this point, but there won't be that many active txns, and,
// even though it'd be an error to have multiple entries for the same txn, we might
// well just get rid of them...
}
else
++it;
}
if (found)
{
semValue++;
idbassert(semValue <= (uint32_t)maxTxns);
condvar.notify_one();
}
else
throw invalid_argument("SessionManagerServer::finishTransaction(): transaction doesn't exist");
}
const TxnID SessionManagerServer::getTxnID(const SID session)
{
TxnID ret;
iterator it;
boost::mutex::scoped_lock lk(mutex);
it = activeTxns.find(session);
if (it != activeTxns.end())
{
ret.id = it->second;
ret.valid = true;
}
return ret;
}
std::shared_ptr<SIDTIDEntry[]> SessionManagerServer::SIDTIDMap(int& len)
{
int j;
std::shared_ptr<SIDTIDEntry[]> ret;
boost::mutex::scoped_lock lk(mutex);
iterator it;
ret.reset(new SIDTIDEntry[activeTxns.size()]);
len = activeTxns.size();
for (it = activeTxns.begin(), j = 0; it != activeTxns.end(); ++it, ++j)
{
ret[j].sessionid = it->first;
ret[j].txnid.id = it->second;
ret[j].txnid.valid = true;
}
return ret;
}
void SessionManagerServer::setSystemState(uint32_t state)
{
boost::mutex::scoped_lock lk(mutex);
systemState |= state;
saveSystemState();
}
void SessionManagerServer::clearSystemState(uint32_t state)
{
boost::mutex::scoped_lock lk(mutex);
systemState &= ~state;
saveSystemState();
}
uint32_t SessionManagerServer::getCpimportJobsCount()
{
std::scoped_lock lk(cpimportMutex);
return activeCpimportJobs.size();
}
uint32_t SessionManagerServer::getTxnCount()
{
boost::mutex::scoped_lock lk(mutex);
return activeTxns.size();
}
void SessionManagerServer::saveSMTxnIDAndState()
{
// caller holds the lock
scoped_ptr<IDBDataFile> txnidfp(IDBDataFile::open(
IDBPolicy::getType(txnidFilename.c_str(), IDBPolicy::WRITEENG), txnidFilename.c_str(), "wb", 0));
if (!txnidfp)
{
perror("SessionManagerServer(): open");
throw runtime_error("SessionManagerServer: Could not open the transaction ID file");
}
int filedata[2];
filedata[0] = _verID;
filedata[1] = _sysCatVerID;
int err = txnidfp->write(filedata, 8);
if (err < 0)
{
perror("SessionManagerServer::newTxnID(): write(verid)");
throw runtime_error("SessionManagerServer::newTxnID(): write(verid) failed");
}
uint32_t lSystemState = systemState;
// We don't save the pending flags, the force flag or the ready flags.
lSystemState &= ~(SS_READY | SS_QUERY_READY | SS_SUSPEND_PENDING | SS_SHUTDOWN_PENDING | SS_FORCE);
err = txnidfp->write(&lSystemState, sizeof(int));
if (err < 0)
{
perror("SessionManagerServer::saveSystemState(): write(systemState)");
throw runtime_error("SessionManagerServer::saveSystemState(): write(systemState) failed");
}
txnidfp->flush();
}
} // namespace BRM