diff --git a/dbcon/mysql/columnstore_info.sql b/dbcon/mysql/columnstore_info.sql index 23a6a0ed6..691ea754d 100644 --- a/dbcon/mysql/columnstore_info.sql +++ b/dbcon/mysql/columnstore_info.sql @@ -82,7 +82,7 @@ DROP PROCEDURE IF EXISTS `compression_ratio` // CREATE PROCEDURE compression_ratio() BEGIN -SELECT CONCAT((SELECT SUM(file_size) FROM information_schema.columnstore_files WHERE compressed_data_size IS NOT NULL) / (SELECT SUM(data_size) FROM information_schema.columnstore_extents) * 100, '%') COMPRESSION_RATIO; +SELECT CONCAT((SELECT SUM(data_size) FROM information_schema.columnstore_extents ce left join information_schema.columnstore_columns cc on ce.object_id = cc.object_id where compression_type='Snappy') / (SELECT SUM(compressed_data_size) FROM information_schema.columnstore_files WHERE compressed_data_size IS NOT NULL), ':1') COMPRESSION_RATIO; END // DELIMITER ; diff --git a/oam/install_scripts/myCnf-include-args.text b/oam/install_scripts/myCnf-include-args.text index bc2c4c74f..40545c7ed 100644 --- a/oam/install_scripts/myCnf-include-args.text +++ b/oam/install_scripts/myCnf-include-args.text @@ -1,7 +1,7 @@ # List of my.cnf arguments that should be checked and saved during upgrade install # infinidb_local_query -log-bin=mysql-bin +log-bin server-id max_length_for_sort_data tmpdir diff --git a/oamapps/postConfigure/mycnfUpgrade.cpp b/oamapps/postConfigure/mycnfUpgrade.cpp index 5fc2f8ddc..afe59e742 100644 --- a/oamapps/postConfigure/mycnfUpgrade.cpp +++ b/oamapps/postConfigure/mycnfUpgrade.cpp @@ -47,7 +47,7 @@ #include #include #include - +#include #include "liboamcpp.h" #include "installdir.h" @@ -118,6 +118,7 @@ int main(int argc, char* argv[]) { includeArg = line; + boost::regex icludeArgRegEx("^#*\\s*" + includeArg + "\\s*="); //see if in my.cnf.rpmsave ifstream mycnfsavefile (mycnfsaveFile.c_str()); char line[200]; @@ -126,9 +127,8 @@ int main(int argc, char* argv[]) while (mycnfsavefile.getline(line, 200)) { oldbuf = line; - string::size_type pos = oldbuf.find(includeArg, 0); - if ( pos != string::npos ) + if ( boost::regex_search(oldbuf.begin(), oldbuf.end(), icludeArgRegEx) ) { //found in my.cnf.rpmsave, check if this is commented out if ( line[0] != '#' ) @@ -144,9 +144,8 @@ int main(int argc, char* argv[]) while (mycnffile.getline(line1, 200)) { newbuf = line1; - string::size_type pos = newbuf.find(includeArg, 0); - if ( pos != string::npos ) + if ( boost::regex_search(newbuf.begin(), newbuf.end(), icludeArgRegEx) ) { newbuf = oldbuf; cout << "Updated argument: " << includeArg << endl; @@ -181,9 +180,9 @@ int main(int argc, char* argv[]) while (mycnffile.getline(line1, 200)) { newbuf = line1; - string::size_type pos = newbuf.find("[mysqld]", 0); + boost::regex mysqldSectionRegEx("\\[mysqld\\]"); - if ( pos != string::npos ) + if ( boost::regex_search(newbuf.begin(), newbuf.end(), mysqldSectionRegEx) ) { lines.push_back(newbuf); newbuf = oldbuf; @@ -206,10 +205,9 @@ int main(int argc, char* argv[]) newFile.close(); close(fd); + break; } } - - break; } } } diff --git a/utils/clusterTester/columnstoreClusterTester.sh b/utils/clusterTester/columnstoreClusterTester.sh index acc39d23a..78fa3f889 100755 --- a/utils/clusterTester/columnstoreClusterTester.sh +++ b/utils/clusterTester/columnstoreClusterTester.sh @@ -978,7 +978,7 @@ checkPackages() fi fi - declare -a DEBIAN9_PKG=("libboost-all-dev" "expect" "libdbi-perl" "perl" "openssl" "file" "sudo" "libreadline5" "rsync" "libsnappy1V5" "net-tools" "libioa1") + declare -a DEBIAN9_PKG=("libboost-all-dev" "expect" "libdbi-perl" "perl" "openssl" "file" "sudo" "libreadline5" "rsync" "libsnappy1V5" "net-tools" "libaio1") if [ "$OS" == "debian9" ]; then if [ ! `which dpkg 2>/dev/null` ] ; then diff --git a/utils/configcpp/xmlparser.cpp b/utils/configcpp/xmlparser.cpp index 93750d9f4..78a4993d1 100644 --- a/utils/configcpp/xmlparser.cpp +++ b/utils/configcpp/xmlparser.cpp @@ -22,6 +22,7 @@ #include "config.h" #include +#include #include #include #include @@ -45,13 +46,17 @@ const string XMLParser::getConfig(const xmlDocPtr doc, const string& section, co while (cur1 != NULL) { - if ((!xmlStrcmp(cur1->name, (const xmlChar*)section.c_str()))) + string cur1name = (const char*)cur1->name; + + if ((boost::iequals(cur1name, section))) { xmlNodePtr cur2 = cur1->xmlChildrenNode; while (cur2 != NULL) { - if ((!xmlStrcmp(cur2->name, (const xmlChar*)name.c_str()))) + string cur2name = (const char*)cur2->name; + + if ((boost::iequals(cur2name, name))) { xmlNodePtr cur3 = cur2->xmlChildrenNode; @@ -85,13 +90,17 @@ void XMLParser::getConfig(const xmlDocPtr doc, const string& section, const stri while (cur1 != NULL) { - if ((!xmlStrcmp(cur1->name, (const xmlChar*)section.c_str()))) + string cur1name = (const char*)cur1->name; + + if ((boost::iequals(cur1name, section))) { xmlNodePtr cur2 = cur1->xmlChildrenNode; while (cur2 != NULL) { - if ((!xmlStrcmp(cur2->name, (const xmlChar*)name.c_str()))) + string cur2name = (const char*)cur2->name; + + if ((boost::iequals(cur2name, name))) { res.clear(); xmlNodePtr cur3 = cur2->xmlChildrenNode; @@ -123,13 +132,17 @@ void XMLParser::setConfig(xmlDocPtr doc, const string& section, const string& na while (cur1 != NULL) { - if (xmlStrcmp(cur1->name, (const xmlChar*)section.c_str()) == 0) + string cur1name = (const char*)cur1->name; + + if (boost::iequals(cur1name, section)) { cur2 = cur1->xmlChildrenNode; while (cur2 != NULL) { - if (xmlStrcmp(cur2->name, (const xmlChar*)name.c_str()) == 0) + string cur2name = (const char*)cur2->name; + + if (boost::iequals(cur2name, name)) { xmlNodePtr cur3 = cur2->xmlChildrenNode; @@ -185,7 +198,9 @@ void XMLParser::delConfig(xmlDocPtr doc, const string& section, const string& na while (cur1 != NULL) { - if ((!xmlStrcmp(cur1->name, (const xmlChar*)section.c_str()))) + string cur1name = (const char*)cur1->name; + + if ((boost::iequals(cur1name, section))) { xmlNodePtr cur2 = cur1->xmlChildrenNode; @@ -193,8 +208,9 @@ void XMLParser::delConfig(xmlDocPtr doc, const string& section, const string& na { xmlNodePtr tmp = cur2; cur2 = cur2->next; + string tmpname = (const char*)tmp->name; - if ((!xmlStrcmp(tmp->name, (const xmlChar*)name.c_str()))) + if ((boost::iequals(tmpname, name))) { xmlUnlinkNode(tmp); xmlFreeNode(tmp); diff --git a/versioning/BRM/extentmap.h b/versioning/BRM/extentmap.h index f65549108..811745145 100644 --- a/versioning/BRM/extentmap.h +++ b/versioning/BRM/extentmap.h @@ -737,8 +737,8 @@ public: * * Function should only be called for System Catalog OIDs, as it assumes * the OID is fully contained on a single DBRoot, returning the first - * DBRoot found. This only makes since for a System Catalog OID, because - * all other column OIDs can span multiple DBRoots. + * DBRoot found. This only makes sence for a System Catalog + * OID, because all other column OIDs can span multiple DBRoots. * * @param oid The system catalog OID * @param dbRoot (out) the DBRoot holding the system catalog OID diff --git a/writeengine/bulk/we_columninfo.cpp b/writeengine/bulk/we_columninfo.cpp index cf57b6c34..9c1a1a9e2 100644 --- a/writeengine/bulk/we_columninfo.cpp +++ b/writeengine/bulk/we_columninfo.cpp @@ -123,7 +123,7 @@ unsigned int compactVarBinary(char* charTmpBuf, int fieldLength) } } - + namespace WriteEngine { @@ -222,7 +222,7 @@ ColumnInfo::ColumnInfo(Log* logger, // fAutoIncMgr = new ColumnAutoIncJob(logger); } } - + //------------------------------------------------------------------------------ // ColumnInfo destructor //------------------------------------------------------------------------------ @@ -263,7 +263,7 @@ void ColumnInfo::clearMemory( ) fDictBlocks.clear(); } - + //------------------------------------------------------------------------------ // If at the start of the job, We have encountered a PM that has no DB file for // this column, or whose HWM extent is disabled; then this function is called @@ -295,7 +295,7 @@ void ColumnInfo::setupDelayedFileCreation( column.compressionType, dbRoot, partition, segment ); } - + //------------------------------------------------------------------------------ // Create a DB file as part of delayed file creation. See setupDelayedFile- // Creation for an explanation. @@ -529,7 +529,7 @@ int ColumnInfo::createDelayedFileIfNeeded( const std::string& tableName ) return rc; } - + //------------------------------------------------------------------------------ // Add an extent for this column. The next segment file in the DBRoot, // partition, segment number rotation will be selected for the extent. @@ -641,7 +641,7 @@ int ColumnInfo::extendColumn( bool saveLBIDForCP ) return rc; } - + //------------------------------------------------------------------------------ // Add a new extent to this column, at the specified DBRoot. Partition may be // used if DBRoot is empty. @@ -804,7 +804,7 @@ int ColumnInfo::extendColumnNewExtent( return NO_ERROR; } - + //------------------------------------------------------------------------------ // Fill out existing partial extent to extent boundary, so that we can resume // inserting rows on an extent boundary basis. This use case should only take @@ -916,7 +916,7 @@ int ColumnInfo::extendColumnOldExtent( return NO_ERROR; } - + //------------------------------------------------------------------------------ // Either add or update the File object, so that it has the updated HWM. // We will access this info to update the HWM in the ExtentMap at the end @@ -970,7 +970,7 @@ void ColumnInfo::addToSegFileList( File& dataFile, HWM hwm ) fSegFileUpdateList.push_back( dataFile ); } } - + //------------------------------------------------------------------------------ // Reset file offset data member attributes when we start working on the next // extent. @@ -998,7 +998,7 @@ int ColumnInfo::resetFileOffsetsNewExtent(const char* /*hdr*/) return NO_ERROR; } - + //------------------------------------------------------------------------------ // Set current size of file in raw (uncompressed) bytes, given the specified // hwm. abbrevFlag indicates whether this is a fixed size abbreviated extent. @@ -1032,7 +1032,7 @@ void ColumnInfo::setFileSize( HWM hwm, int abbrevFlag ) } } } - + //------------------------------------------------------------------------------ // If we are dealing with the first extent in the first segment file for this // column, and the segment file is still equal to 256K rows, then we set the @@ -1064,7 +1064,7 @@ void ColumnInfo::setAbbrevExtentCheck( ) } } } - + //------------------------------------------------------------------------------ // If this is an abbreviated extent, we expand the extent to a full extent on // disk, by initializing the necessary number of remaining blocks. @@ -1144,7 +1144,7 @@ int ColumnInfo::expandAbbrevExtent( bool bRetainFilePos ) return NO_ERROR; } - + //------------------------------------------------------------------------------ // Close the current Column file. //------------------------------------------------------------------------------ @@ -1158,7 +1158,7 @@ int ColumnInfo::closeColumnFile(bool /*bCompletingExtent*/, bool /*bAbort*/) return NO_ERROR; } - + //------------------------------------------------------------------------------ // Initialize fLastInputRowInCurrentExtent used in detecting when a Read Buffer // is crossing an extent boundary, so that we can accurately track the min/max @@ -1207,7 +1207,7 @@ void ColumnInfo::lastInputRowInExtentInit( bool bIsNewExtent ) bIsNewExtent ); } } - + //------------------------------------------------------------------------------ // Increment fLastRIDInExtent to the end of the next extent. //------------------------------------------------------------------------------ @@ -1215,7 +1215,7 @@ void ColumnInfo::lastInputRowInExtentInc( ) { fLastInputRowInCurrentExtent += fRowsPerExtent; } - + //------------------------------------------------------------------------------ // Parsing is complete for this column. Flush pending data. Close the current // segment file, and corresponding dictionary store file (if applicable). Also @@ -1281,7 +1281,7 @@ int ColumnInfo::finishParsing( ) return NO_ERROR; } - + //------------------------------------------------------------------------------ // Store updated column information in BRMReporter for this column at EOJ; // so that Extent Map CP information and HWM's can be updated. @@ -1306,7 +1306,7 @@ void ColumnInfo::getBRMUpdateInfo( BRMReporter& brmReporter ) if (entriesAdded > 0) getCPInfoForBRM( brmReporter ); } - + //------------------------------------------------------------------------------ // Get updated Casual Partition (CP) information for BRM for this column at EOJ. //------------------------------------------------------------------------------ @@ -1314,7 +1314,7 @@ void ColumnInfo::getCPInfoForBRM( BRMReporter& brmReporter ) { fColExtInf->getCPInfoForBRM(column, brmReporter); } - + //------------------------------------------------------------------------------ // Get updated HWM information for BRM for this column at EOJ. // Returns count of the number of HWM entries added to the BRMReporter. @@ -1388,7 +1388,7 @@ int ColumnInfo::getHWMInfoForBRM( BRMReporter& brmReporter ) return entriesAdded; } - + //------------------------------------------------------------------------------ // Setup initial extent we will begin loading at start of import. // DBRoot, partition, segment, etc for the starting extent are specified. @@ -1516,7 +1516,7 @@ int ColumnInfo::setupInitialColumnExtent( return NO_ERROR; } - + //------------------------------------------------------------------------------ // Prepare the initial column segment file for import. //------------------------------------------------------------------------------ @@ -1579,7 +1579,7 @@ int ColumnInfo::setupInitialColumnFile( HWM oldHwm, HWM hwm ) return NO_ERROR; } - + //------------------------------------------------------------------------------ // Update the number of bytes in the file, and the free space still remaining. //------------------------------------------------------------------------------ @@ -1600,7 +1600,7 @@ bool ColumnInfo::isFileComplete() const return false; } - + //------------------------------------------------------------------------------ // Initialize last used auto-increment value from the current "next" // auto-increment value taken from the system catalog (or BRM). @@ -1633,7 +1633,7 @@ int ColumnInfo::finishAutoInc( ) return rc; } - + //------------------------------------------------------------------------------ // Get current dbroot, partition, segment, and HWM for this column. // @@ -1656,7 +1656,7 @@ void ColumnInfo::getSegFileInfo( DBRootExtentInfo& fileInfo ) else fileInfo.fLocalHwm = 0; } - + //------------------------------------------------------------------------------ // Open a new or existing Dictionary store file based on the DBRoot, // partition, and segment settings in curCol.dataFile. @@ -1805,7 +1805,7 @@ int ColumnInfo::openDctnryStore( bool bMustExist ) return rc; } - + //------------------------------------------------------------------------------ // Close the current Dictionary store file. //------------------------------------------------------------------------------ @@ -1837,7 +1837,7 @@ int ColumnInfo::closeDctnryStore(bool bAbort) return rc; } - + //------------------------------------------------------------------------------ // Update dictionary store file with specified strings, and return the assigned // tokens (tokenbuf) to be stored in the corresponding column token file. @@ -1903,7 +1903,7 @@ int ColumnInfo::updateDctnryStore(char* buf, return NO_ERROR; } - + //------------------------------------------------------------------------------ // No action necessary for uncompressed dictionary files //------------------------------------------------------------------------------ @@ -1924,7 +1924,7 @@ const { return NO_ERROR; } - + //------------------------------------------------------------------------------ // utility to convert a Status enumeration to a string //------------------------------------------------------------------------------ diff --git a/writeengine/shared/we_fileop.cpp b/writeengine/shared/we_fileop.cpp index c4fea1c16..7a2a2456b 100644 --- a/writeengine/shared/we_fileop.cpp +++ b/writeengine/shared/we_fileop.cpp @@ -605,8 +605,7 @@ int FileOp::extendFile( return ERR_FILE_NOT_EXIST; } - pFile = openFile( oid, dbRoot, partition, segment, - segFile, "r+b" );//old file + pFile = openFile( oid, dbRoot, partition, segment, segFile, "r+b" );//old file if (pFile == 0) { @@ -626,7 +625,7 @@ int FileOp::extendFile( if ( isDebug(DEBUG_1) && getLogger() ) { std::ostringstream oss; - oss << "Opening existing column file" << + oss << "Opening existing column file (extendFile)" << ": OID-" << oid << "; DBRoot-" << dbRoot << "; part-" << partition << @@ -664,19 +663,60 @@ int FileOp::extendFile( ostringstream oss; oss << "oid: " << oid << " with path " << segFile << "; new extent fbo " << hwm << "; number of " - "compressed chunks " << ptrCount; + "compressed chunks " << ptrCount << + "; chunkIndex " << chunkIndex; logging::Message::Args args; args.add("compressed"); args.add(oss.str()); SimpleSysLog::instance()->logMsg(args, logging::LOG_TYPE_ERROR, logging::M0103); - return ERR_FILE_NEW_EXTENT_FBO; + + // Expand the partial extent to full with emptyVal + // Since fillCompColumnExtentEmptyChunks() messes with the + // file on disk, we need to close it and reopen after or + // the cache isn't updated. + if ((pFile)) + closeFile( pFile ); + + pFile = NULL; + string failedTask; // For return error message, if any. + rc = FileOp::fillCompColumnExtentEmptyChunks(oid, + width, + emptyVal, + dbRoot, + partition, + segment, + hwm, + segFile, + failedTask); + + if (rc != NO_ERROR) + { + if (getLogger()) + { + std::ostringstream oss; + oss << "FileOp::extendFile: error padding partial compressed extent for " << + "column OID-" << oid << + "; DBRoot-" << dbRoot << + "; part-" << partition << + "; seg-" << segment << + "; hwm-" << hwm << + " " << failedTask.c_str(); + getLogger()->logMsg( oss.str(), rc, MSGLVL_CRITICAL ); + } + + return rc; + } + + pFile = openFile( oid, dbRoot, partition, segment, segFile, "r+b" ); // modified file } + // Get the latest file header for the caller. If a partial extent was filled out, + // this will be different than when we first read the headers. if (hdrs) { - memcpy(hdrs, hdrsIn, sizeof(hdrsIn) ); + RETURN_ON_ERROR( readHeaders(pFile, hdrs) ); } } else @@ -706,7 +746,28 @@ int FileOp::extendFile( SimpleSysLog::instance()->logMsg(args, logging::LOG_TYPE_ERROR, logging::M0103); - return ERR_FILE_NEW_EXTENT_FBO; + // Expand the partial extent to full with emptyVal + // This generally won't ever happen, as uncompressed files + // are created with full extents. + rc = FileOp::expandAbbrevColumnExtent( pFile, dbRoot, + emptyVal, width); + + if (rc != NO_ERROR) + { + if (getLogger()) + { + std::ostringstream oss; + oss << "FileOp::extendFile: error padding partial uncompressed extent for " << + "column OID-" << oid << + "; DBRoot-" << dbRoot << + "; part-" << partition << + "; seg-" << segment << + "; hwm-" << hwm; + getLogger()->logMsg( oss.str(), rc, MSGLVL_CRITICAL ); + } + + return rc; + } } } } @@ -2726,10 +2787,10 @@ int FileOp::expandAbbrevColumnExtent( } // Add blocks to turn the abbreviated extent into a full extent. - int rc = initColumnExtent(pFile, dbRoot, blksToAdd, emptyVal, width, - false, // existing file - true, // expand existing extent - false); // n/a since not adding new extent + int rc = FileOp::initColumnExtent(pFile, dbRoot, blksToAdd, emptyVal, width, + false, // existing file + true, // expand existing extent + false); // n/a since not adding new extent return rc; } diff --git a/writeengine/shared/we_log.h b/writeengine/shared/we_log.h index a3aef6ad6..bc226789d 100644 --- a/writeengine/shared/we_log.h +++ b/writeengine/shared/we_log.h @@ -48,7 +48,7 @@ namespace WriteEngine const std::string MSG_LEVEL_STR[] = { "INFO", - "INFO", + "INFO2", "WARN", "ERR ", "CRIT" diff --git a/writeengine/splitter/we_sdhandler.cpp b/writeengine/splitter/we_sdhandler.cpp index 7baf14a3a..084655924 100644 --- a/writeengine/splitter/we_sdhandler.cpp +++ b/writeengine/splitter/we_sdhandler.cpp @@ -1110,7 +1110,7 @@ void WESDHandler::onNakResponse(int PmId) // consume too much CPU void WESDHandler::onEodResponse(int PmId) { - if (getDebugLvl()) + if (getDebugLvl() > 2) cout << "Received a EOD from " << PmId << endl; if (fRef.fCmdArgs.getMode() == 0) diff --git a/writeengine/wrapper/we_colopcompress.cpp b/writeengine/wrapper/we_colopcompress.cpp index 932a1d688..cb4fb05dd 100644 --- a/writeengine/wrapper/we_colopcompress.cpp +++ b/writeengine/wrapper/we_colopcompress.cpp @@ -194,7 +194,14 @@ int ColumnOpCompress1::expandAbbrevColumnExtent( IDBDataFile* pFile, uint16_t dbRoot, uint64_t emptyVal, int width) { // update the uncompressed initial chunk to full chunk - RETURN_ON_ERROR(m_chunkManager->expandAbbrevColumnExtent(pFile, emptyVal, width)); + int rc = m_chunkManager->expandAbbrevColumnExtent(pFile, emptyVal, width); + + // ERR_COMP_FILE_NOT_FOUND is acceptable here. It just means that the + // file hasn't been loaded into the chunk manager yet. No big deal. + if (rc != NO_ERROR && rc != ERR_COMP_FILE_NOT_FOUND) + { + return rc; + } // let the base to physically expand extent. return FileOp::expandAbbrevColumnExtent(pFile, dbRoot, emptyVal, width);