database/mariadb-columnstore-engine
1
0
Fork 0
mirror of https://github.com/mariadb-corporation/mariadb-columnstore-engine.git synced 2025-08-07 03:22:57 +03:00
Code Activity

Add data redundancy feature.

Browse Source
This commit is contained in:
Ben Thompson
2017-07-20 10:01:39 -05:00
parent 47fb014be3
commit 7ec2a244ef
9 changed files with 917 additions and 217 deletions
Download Patch File Download Diff File Expand all files Collapse all files

325
oam/oamcpp/liboamcpp.cpp
View File

@@ -6899,11 +6899,12 @@ namespace oam
if ( GlusterConfig == "y")
{
try {
string errmsg;
int ret = glusterctl(oam::GLUSTER_DELETE, itoa(dbrootID), errmsg, errmsg);
string errmsg1;
string errmsg2;
int ret = glusterctl(oam::GLUSTER_DELETE, itoa(dbrootID), errmsg1, errmsg2);
if ( ret != 0 )
{
cerr << "FAILURE: Error deleting gluster dbroot# " + itoa(dbrootID) + ", error: " + errmsg << endl;
cerr << "FAILURE: Error deleting gluster dbroot# " + itoa(dbrootID) + ", error: " + errmsg1 << endl;
exceptionControl("removeDbroot", API_FAILURE);
}
}
@@ -8098,13 +8099,12 @@ namespace oam
return;
}
/******************************************************************************************
* @brief glusterctl
*
* purpose: gluster control and glusteradd
*
* commands: status - Used to check status of kuster and disk good/bad
* commands: status - Used to check status of gluster and disk good/bad
* returns
* NOTINSTALLED
* OK
@@ -8156,40 +8156,42 @@ namespace oam
#ifndef _MSC_VER
int user;
user = getuid();
string glustercmd = InstallDir + "/bin/glusterctl";
string glustercmd = "gluster ";
if (access(glustercmd.c_str(), X_OK) != 0 )
exceptionControl("glusterctl", API_DISABLED);
if (user!=0)
{
glustercmd = "sudo " + glustercmd;
}
errmsg = "";
switch ( command ) {
case (oam::GLUSTER_STATUS):
{
glustercmd = glustercmd + " status > /tmp/gluster_status.log 2>&1";
string command = glustercmd + "volume status";
int ret;
ret = system(glustercmd.c_str());
if ( WEXITSTATUS(ret) == 0 )
return 0;
ret = checkGlusterLog("/tmp/gluster_status.log", errmsg);
return ret;
char buffer[128];
string result = "";
FILE* pipe = popen(command.c_str(), "r");
if (!pipe) exceptionControl("GLUSTER_STATUS", API_FAILURE,"popen() failed");
try {
while (!feof(pipe)) {
if (fgets(buffer, 128, pipe) != NULL)
result += buffer;
}
} catch (...) {
exceptionControl("GLUSTER_STATUS", API_FAILURE);
throw;
}
pclose(pipe);
argument2 = result;
return 0;
}
case (oam::GLUSTER_SETDDC):
{
string copies = argument1;
writeLog("glusterctl: GLUSTER_SETDDC: " + copies, LOG_TYPE_DEBUG );
glustercmd = glustercmd + " setddc " + copies + " > /tmp/gluster_setddc.log 2>&1";
int ret;
ret = system(glustercmd.c_str());
if ( WEXITSTATUS(ret) == 0 )
return 0;
ret = checkGlusterLog("/tmp/gluster_setddc.log", errmsg);
return ret;
// The way this was implemented it doesn't seem possible to actually change the value..
return 0;
}
case (oam::GLUSTER_ASSIGN):
@@ -8202,18 +8204,7 @@ namespace oam
if (returnStatus != API_SUCCESS)
exceptionControl("GLUSTER_ASSIGN", returnStatus);
/*
writeLog("glusterctl call: GLUSTER_ASSIGN: dbroot = " + dbrootID + " pm = " + pmID, LOG_TYPE_DEBUG );
glustercmd = glustercmd + " assign " + dbrootID + " " + pmID + " > /tmp/gluster_assign.log 2>&1";
int ret;
ret = system(glustercmd.c_str());
writeLog("glusterctl return: GLUSTER_ASSIGN: dbroot = " + dbrootID + " pm = " + pmID, LOG_TYPE_DEBUG );
if ( WEXITSTATUS(ret) == 0 )
return 0;
ret = checkGlusterLog("/tmp/gluster_assign.log", errmsg);
return ret;
*/
break;
}
@@ -8228,84 +8219,188 @@ namespace oam
if (returnStatus != API_SUCCESS)
exceptionControl("GLUSTER_UNASSIGN", returnStatus);
/* writeLog("glusterctl: GLUSTER_UNASSIGN: dbroot = " + dbrootID + " pm = " + pmID, LOG_TYPE_DEBUG );
glustercmd = glustercmd + " unassign " + dbrootID + " " + pmID + " > /tmp/gluster_unassign.log 2>&1";
int ret;
ret = system(glustercmd.c_str());
if ( WEXITSTATUS(ret) == 0 )
return 0;
ret = checkGlusterLog("/tmp/gluster_unassign.log", errmsg);
return ret;
*/
break;
}
case (oam::GLUSTER_WHOHAS):
{
// Script returned a string of space separated PMIDs everywhere this is called it expects that return form.
// When configuring or adding Modules we need to update DataRedundancyConfig section of configuration file.
Config* sysConfig = Config::makeConfig(CalpontConfigFile.c_str());
string dbrootID = argument1;
string msg;
for ( int retry = 0 ; retry < 5 ; retry++ )
string moduleDBRootIDPMs = "DBRoot" + dbrootID + "PMs";
string msg = sysConfig->getConfig("DataRedundancyConfig",moduleDBRootIDPMs);
if (msg.empty())
{
writeLog("glusterctl: GLUSTER_WHOHAS for dbroot : " + dbrootID, LOG_TYPE_DEBUG );
glustercmd = glustercmd + " whohas " + dbrootID + " > /tmp/gluster_howhas.log 2>&1";
system(glustercmd.c_str());
int ret;
ret = checkGlusterLog("/tmp/gluster_howhas.log", msg);
if ( ret == 0 )
{ // OK return, get pm list
if ( msg.empty() )
{
writeLog("glusterctl: GLUSTER_WHOHAS: empty pm list", LOG_TYPE_ERROR );
exceptionControl("glusterctl", API_FAILURE);
}
else
{
writeLog("glusterctl: GLUSTER_WHOHAS: pm list = " + msg, LOG_TYPE_DEBUG );
argument2 = msg;
return 0;
}
}
// retry failure
writeLog("glusterctl: GLUSTER_WHOHAS: failure, retrying (restarting gluster) " + msg, LOG_TYPE_ERROR );
string cmd = "service glusterd restart > /dev/null 2>&1";
if (user != 0)
cmd = "sudo " + cmd;
system(cmd.c_str());
cmd = "systemctrl restart glusterd > /dev/null 2>&1";
if (user != 0)
cmd = "sudo " + cmd;
system(cmd.c_str());
sleep(1);
exceptionControl("glusterctl", API_FAILURE);
}
argument2 = msg;
return 0;
break;
}
case (oam::GLUSTER_ADD):
{
string pmID = argument1;
string dbrootID = argument2;
int pmID = atoi(argument1.c_str());
int dbrootID = atoi(argument2.c_str());
string command = "";
writeLog("glusterctl: GLUSTER_ADD: dbroot = " + argument2 + " pm = " + argument1, LOG_TYPE_DEBUG );
string glustercmd = InstallDir + "/bin/glusteradd";
writeLog("glusterctl: GLUSTER_ADD: dbroot = " + dbrootID + " pm = " + pmID, LOG_TYPE_DEBUG );
glustercmd = glustercmd + " " + pmID + " " + dbrootID + " > /tmp/gluster_add.log 2>&1";
int ret;
//writeLog("glusterctl: cmd = " + glustercmd, LOG_TYPE_DEBUG );
ret = system(glustercmd.c_str());
if ( WEXITSTATUS(ret) == 0 )
return 0;
Config* sysConfig = Config::makeConfig();
ModuleTypeConfig moduletypeconfig;
getSystemConfig("pm", moduletypeconfig);
int dataRedundancyCopies;
int dbrootCount;
getSystemConfig("DataRedundancyCopies",dataRedundancyCopies);
getSystemConfig("DBRootCount",dbrootCount);
ret = checkGlusterLog("/tmp/gluster_add.log", errmsg);
return ret;
int numberPMs = moduletypeconfig.ModuleCount;
int numberNewPMs = numberPMs - pmID + 1;
int numberNewDBRoots = (dbrootCount - dbrootID) + 1;
int numberDBRootsPerPM = numberNewDBRoots/numberNewPMs;
std::vector<int> dbrootPms[dbrootCount];
DataRedundancyConfig DataRedundancyConfigs[numberPMs];
int startDBRootID = dbrootID;
for (int pm=(pmID-1); pm < numberPMs; pm++,startDBRootID++)
{
DataRedundancyConfigs[pm].pmID = (pm+1);
string moduleHostName = "ModuleHostName" + itoa(pm+1) + "-1-3";
string moduleIpAddr = "ModuleIPAddr" + itoa(pm+1) + "-1-3";
DataRedundancyConfigs[pm].pmHostname = sysConfig->getConfig("DataRedundancyConfig",moduleHostName);
DataRedundancyConfigs[pm].pmIpAddr = sysConfig->getConfig("DataRedundancyConfig",moduleIpAddr);
if (DataRedundancyConfigs[pm].pmHostname.empty())
{
DataRedundancyConfigs[pm].pmHostname = sysConfig->getConfig("SystemModuleConfig",moduleHostName);
}
if (DataRedundancyConfigs[pm].pmIpAddr.empty())
{
DataRedundancyConfigs[pm].pmIpAddr = sysConfig->getConfig("SystemModuleConfig",moduleIpAddr);
}
int nextPM = (pm+1);
int dbrootCopy = startDBRootID;
if (nextPM >= numberPMs)
{
nextPM=(pmID-1);
}
for ( int i=0; i<numberDBRootsPerPM; i++)
{
DataRedundancyConfigs[pm].dbrootCopies.push_back(dbrootCopy);
for ( int copies = 0; copies < (dataRedundancyCopies-1); copies++)
{
DataRedundancyConfigs[nextPM].dbrootCopies.push_back(dbrootCopy);
nextPM++;
if (nextPM >= numberPMs)
{
nextPM=(pmID-1);
}
if (nextPM == pm)
{
nextPM++;
}
if (nextPM >= numberPMs)
{
nextPM=(pmID-1);
}
}
dbrootCopy += numberNewPMs;
}
}
for (int db=(dbrootID-1); db < dbrootCount; db++)
{
int newDBrootID = db + 1;
string dbrootIDPMs = "";
string moduleDBRootIDPMs = "DBRoot" + itoa(newDBrootID) + "PMs";
for (int pm=(pmID-1); pm < numberPMs; pm++)
{
if(find(DataRedundancyConfigs[pm].dbrootCopies.begin(),DataRedundancyConfigs[pm].dbrootCopies.end(),newDBrootID) != DataRedundancyConfigs[pm].dbrootCopies.end())
{
dbrootPms[db].push_back(DataRedundancyConfigs[pm].pmID);
dbrootIDPMs += itoa(DataRedundancyConfigs[pm].pmID);
dbrootIDPMs += " ";
}
}
// Store to config and distribute so that GLUSTER_WHOHAS will work
sysConfig->setConfig("DataRedundancyConfig", moduleDBRootIDPMs, dbrootIDPMs);
}
for (int pm=(pmID-1); pm < numberPMs; pm++)
{
command = glustercmd + "peer probe " + DataRedundancyConfigs[pm].pmIpAddr;
int status = system(command.c_str());
if (WEXITSTATUS(status) != 0 )
{
writeLog("ERROR: command failed: ",LOG_TYPE_DEBUG);
exceptionControl("GLUSTER_ADD", API_FAILURE);
}
}
sleep(5);
command = glustercmd + "peer status ";
int status = system(command.c_str());
if (WEXITSTATUS(status) != 0 )
{
cout << "ERROR: command failed: " << command << endl;
exit(1);
}
//Need to wait since peer probe success does not always mean it is ready for volume create command
sleep(10);
for (int db=(dbrootID-1); db < dbrootCount; db++)
{
int newDbrootID = db + 1;
command = glustercmd + "volume create dbroot" + itoa(newDbrootID) + " transport tcp replica " + itoa(dataRedundancyCopies) + " ";
vector<int>::iterator dbrootPmIter = dbrootPms[db].begin();
for (; dbrootPmIter < dbrootPms[db].end(); dbrootPmIter++ )
{
int pm = (*dbrootPmIter) - 1;
command += DataRedundancyConfigs[pm].pmIpAddr + ":" + InstallDir +"/gluster/brick" + itoa(newDbrootID) + " ";
}
command += "force";
int status = system(command.c_str());
if (WEXITSTATUS(status) != 0 )
{
writeLog("ERROR: command failed: " + command,LOG_TYPE_DEBUG);
exceptionControl("GLUSTER_ADD", API_FAILURE);
}
command = glustercmd + "volume start dbroot" + itoa(newDbrootID);
status = system(command.c_str());
if (WEXITSTATUS(status) != 0 )
{
writeLog("ERROR: command failed: ",LOG_TYPE_DEBUG);
exceptionControl("GLUSTER_ADD", API_FAILURE);
}
}
try
{
sysConfig->write();
}
catch(...)
{
exceptionControl("sysConfig->write", API_FAILURE);
}
distributeConfigFile("system");
for (int pm=(pmID-1); pm < numberPMs; pm++)
{
for (int i=0; i<numberDBRootsPerPM; i++)
{
string ModuleDBRootID = "ModuleDBRootID"+ itoa(pm+1) + "-" + itoa(i+1) +"-3";
string dbr = sysConfig->getConfig("SystemModuleConfig",ModuleDBRootID);
string command = "" + DataRedundancyConfigs[pm].pmIpAddr +
":/dbroot" + dbr + " " + InstallDir + "/data" + dbr +
" glusterfs defaults,direct-io-mode=enable 00";
string toPM = "pm" + itoa(pm+1);
distributeFstabUpdates(command,toPM);
}
}
break;
}
@@ -8313,18 +8408,27 @@ namespace oam
case (oam::GLUSTER_DELETE):
{
string dbrootID = argument1;
string command = "";
int status;
writeLog("glusterctl: GLUSTER_DELETE: dbroot = " + dbrootID, LOG_TYPE_DEBUG );
glustercmd = glustercmd + " delete " + dbrootID + " > /tmp/gluster_delete.log 2>&1";
int ret;
//writeLog("glusterctl: cmd = " + glustercmd, LOG_TYPE_DEBUG );
ret = system(glustercmd.c_str());
if ( WEXITSTATUS(ret) == 0 )
return 0;
ret = checkGlusterLog("/tmp/gluster_delete.log", errmsg);
return ret;
command = glustercmd + "volume stop dbroot" + dbrootID;
status = system(command.c_str());
if (WEXITSTATUS(status) != 0 )
{
writeLog("ERROR: command failed: ",LOG_TYPE_DEBUG);
exceptionControl("GLUSTER_DELETE", API_FAILURE);
}
command = glustercmd + "volume delete dbroot" + dbrootID;
status = system(command.c_str());
if (WEXITSTATUS(status) != 0 )
{
writeLog("ERROR: command failed: ",LOG_TYPE_DEBUG);
exceptionControl("GLUSTER_DELETE", API_FAILURE);
}
break;
}
@@ -8649,7 +8753,6 @@ namespace oam
return entry;
}
/***************************************************************************
* PRIVATE FUNCTIONS
***************************************************************************/