You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@trafodion.apache.org by db...@apache.org on 2018/03/01 21:08:01 UTC
[3/6] trafodion git commit: Code review fixes.
Code review fixes.
Project: http://git-wip-us.apache.org/repos/asf/trafodion/repo
Commit: http://git-wip-us.apache.org/repos/asf/trafodion/commit/3d7855b6
Tree: http://git-wip-us.apache.org/repos/asf/trafodion/tree/3d7855b6
Diff: http://git-wip-us.apache.org/repos/asf/trafodion/diff/3d7855b6
Branch: refs/heads/master
Commit: 3d7855b6f64733ad3776fc421cb598883acbb6bf
Parents: bded0e8
Author: Zalo Correa <za...@esgyn.com>
Authored: Wed Feb 28 15:23:31 2018 -0800
Committer: Zalo Correa <za...@esgyn.com>
Committed: Wed Feb 28 15:23:31 2018 -0800
----------------------------------------------------------------------
core/sqf/monitor/linux/cluster.cxx | 128 ++++++++++++++++----------------
core/sqf/monitor/linux/cluster.h | 25 +++----
core/sqf/monitor/linux/pnode.cxx | 1 -
core/sqf/monitor/linux/tmsync.cxx | 10 +--
core/sqf/monitor/linux/zclient.cxx | 8 +-
5 files changed, 87 insertions(+), 85 deletions(-)
----------------------------------------------------------------------
http://git-wip-us.apache.org/repos/asf/trafodion/blob/3d7855b6/core/sqf/monitor/linux/cluster.cxx
----------------------------------------------------------------------
diff --git a/core/sqf/monitor/linux/cluster.cxx b/core/sqf/monitor/linux/cluster.cxx
index d1b3e91..83ea923 100644
--- a/core/sqf/monitor/linux/cluster.cxx
+++ b/core/sqf/monitor/linux/cluster.cxx
@@ -380,45 +380,46 @@ void CCluster::AssignMonitorLeader( int pnid )
int i = 0;
int rc = 0;
- int lMonitorLeaderPNid = MonitorLeaderPNid;
+ int monitorLeaderPNid = monitorLeaderPNid_;
CNode *node = NULL;
- if (MonitorLeaderPNid != pnid)
+ if (monitorLeaderPNid_ != pnid)
{
if (trace_settings & (TRACE_INIT | TRACE_RECOVERY | TRACE_REQUEST | TRACE_SYNC | TRACE_TMSYNC))
{
trace_printf( "%s@%d" " - (MasterMonitor) returning, pnid %d != monitorLead %d\n"
- , method_name, __LINE__, pnid, MonitorLeaderPNid );
+ , method_name, __LINE__, pnid, monitorLeaderPNid_ );
}
- return;
+ TRACE_EXIT;
+ return;
}
if (trace_settings & (TRACE_INIT | TRACE_RECOVERY | TRACE_REQUEST | TRACE_SYNC | TRACE_TMSYNC))
{
trace_printf( "%s@%d" " - (MasterMonitor) Node " "%d" " MonitorLeader failed!\n"
- , method_name, __LINE__, MonitorLeaderPNid );
+ , method_name, __LINE__, monitorLeaderPNid_ );
}
for (i=0; i<GetConfigPNodesMax(); i++)
{
- lMonitorLeaderPNid++;
+ monitorLeaderPNid++;
- if (lMonitorLeaderPNid == GetConfigPNodesMax())
+ if (monitorLeaderPNid == GetConfigPNodesMax())
{
- lMonitorLeaderPNid = 0; // restart with nid 0
+ monitorLeaderPNid = 0; // restart with nid 0
}
- if (lMonitorLeaderPNid == pnid)
+ if (monitorLeaderPNid == pnid)
{
continue; // this is the node that is going down, skip it
}
- if (Node[lMonitorLeaderPNid] == NULL)
+ if (Node[monitorLeaderPNid] == NULL)
{
continue;
}
- node = Node[lMonitorLeaderPNid];
+ node = Node[monitorLeaderPNid];
if (trace_settings & (TRACE_INIT | TRACE_RECOVERY | TRACE_REQUEST | TRACE_SYNC | TRACE_TMSYNC))
{
@@ -438,11 +439,11 @@ void CCluster::AssignMonitorLeader( int pnid )
continue; // skip this node for any of the above reasons
}
- MonitorLeaderPNid = node->GetPNid();
+ monitorLeaderPNid_ = node->GetPNid();
if (trace_settings & (TRACE_INIT | TRACE_RECOVERY | TRACE_REQUEST | TRACE_SYNC | TRACE_TMSYNC))
{
- trace_printf("%s@%d" " - Node " "%d" " is the new MonitorLeaderPNid." "\n", method_name, __LINE__, MonitorLeaderPNid);
+ trace_printf("%s@%d" " - Node " "%d" " is the new monitorLeaderPNid_." "\n", method_name, __LINE__, monitorLeaderPNid_);
}
if (ZClientEnabled)
@@ -483,7 +484,7 @@ void CCluster::AssignMonitorLeader( int pnid )
TRACE_EXIT;
}
-// Assigns a new TMLeader if given pnid is same as TmLeaderNid
+// Assigns a new TMLeader if given pnid is same as tmLeaderNid_
// TmLeader is a logical node num.
// pnid has gone down, so if that node was previously the TM leader, a new one needs to be chosen.
void CCluster::AssignTmLeader( int pnid, bool checkProcess )
@@ -495,15 +496,15 @@ void CCluster::AssignTmLeader( int pnid, bool checkProcess )
CNode *node = NULL;
CProcess *process = NULL;
- int TmLeaderPNid = LNode[TmLeaderNid]->GetNode()->GetPNid();
+ int TmLeaderPNid = LNode[tmLeaderNid_]->GetNode()->GetPNid();
if (TmLeaderPNid != pnid)
{
- node = LNode[TmLeaderNid]->GetNode();
+ node = LNode[tmLeaderNid_]->GetNode();
if (checkProcess)
{
- process = LNode[TmLeaderNid]->GetProcessLByType( ProcessType_DTM );
+ process = LNode[tmLeaderNid_]->GetProcessLByType( ProcessType_DTM );
if (process)
{
if (trace_settings & (TRACE_INIT | TRACE_RECOVERY | TRACE_REQUEST | TRACE_SYNC | TRACE_TMSYNC))
@@ -544,7 +545,7 @@ void CCluster::AssignTmLeader( int pnid, bool checkProcess )
if (trace_settings & (TRACE_INIT | TRACE_RECOVERY | TRACE_REQUEST | TRACE_SYNC | TRACE_TMSYNC))
{
trace_printf( "%s@%d" " - Node " "%d" " TmLeader failed! (checkProcess=%d)\n"
- , method_name, __LINE__, TmLeaderNid, checkProcess );
+ , method_name, __LINE__, tmLeaderNid_, checkProcess );
}
for (i=0; i<GetConfigPNodesMax(); i++)
@@ -586,11 +587,11 @@ void CCluster::AssignTmLeader( int pnid, bool checkProcess )
continue; // skip this node for any of the above reasons
}
- TmLeaderNid = node->GetFirstLNode()->GetNid();
+ tmLeaderNid_ = node->GetFirstLNode()->GetNid();
if (checkProcess)
{
- process = LNode[TmLeaderNid]->GetProcessLByType( ProcessType_DTM );
+ process = LNode[tmLeaderNid_]->GetProcessLByType( ProcessType_DTM );
if (!process)
{
continue; // skip this node no DTM process exists
@@ -599,7 +600,7 @@ void CCluster::AssignTmLeader( int pnid, bool checkProcess )
if (trace_settings & (TRACE_INIT | TRACE_RECOVERY | TRACE_REQUEST | TRACE_SYNC | TRACE_TMSYNC))
{
- trace_printf("%s@%d" " - Node " "%d" " is the new TmLeader." "\n", method_name, __LINE__, TmLeaderNid);
+ trace_printf("%s@%d" " - Node " "%d" " is the new TmLeader." "\n", method_name, __LINE__, tmLeaderNid_);
}
break;
@@ -618,14 +619,13 @@ CCluster::CCluster (void)
,epollFD_(-1),
Node (NULL),
LNode (NULL),
- TmSyncPNid (-1),
- CurNodes (0),
- CurProcs (0),
+ tmSyncPNid_ (-1),
+ currentNodes_ (0),
configPNodesCount_ (-1),
configPNodesMax_ (-1),
- NodeMap (NULL),
- TmLeaderNid (-1),
- MonitorLeaderPNid (-1),
+ nodeMap_ (NULL),
+ tmLeaderNid_ (-1),
+ monitorLeaderPNid_ (-1),
tmReadyCount_(0),
minRecvCount_(4096),
recvBuffer_(NULL),
@@ -795,10 +795,10 @@ CCluster::~CCluster (void)
delete [] otherMonRank_;
delete [] socks_;
delete [] sockPorts_;
- if (NodeMap)
+ if (nodeMap_)
{
- delete [] NodeMap;
- NodeMap = NULL;
+ delete [] nodeMap_;
+ nodeMap_ = NULL;
}
delete [] recvBuffer2_;
@@ -2416,7 +2416,7 @@ void CCluster::HandleOtherNodeMsg (struct internal_msg_def *recv_msg,
// Begin a Slave Sync Start
if (trace_settings & (TRACE_SYNC | TRACE_TMSYNC))
trace_printf("%s@%d - Slave Sync Start on Node %s (pnid=%d)\n", method_name, __LINE__, Node[pnid]->GetName(), pnid);
- TmSyncPNid = pnid;
+ tmSyncPNid_ = pnid;
Node[pnid]->SetTmSyncState( recv_msg->u.sync.state );
if (trace_settings & (TRACE_SYNC | TRACE_TMSYNC))
{
@@ -2430,12 +2430,12 @@ void CCluster::HandleOtherNodeMsg (struct internal_msg_def *recv_msg,
trace_printf("%s@%d - Sync State Collision! Node %s (pnid=%d) TmSyncState=(%d)(%s)\n", method_name, __LINE__, MyNode->GetName(), MyPNID, MyNode->GetTmSyncState(), SyncStateString( MyNode->GetTmSyncState()) );
if ( MyNode->GetTmSyncState() == SyncState_Continue )
{
- if ( pnid > TmSyncPNid )
+ if ( pnid > tmSyncPNid_ )
// highest node id will continue
{
// They take priority ... we abort
if (trace_settings & (TRACE_SYNC | TRACE_TMSYNC))
- trace_printf("%s@%d - Aborting Slave Sync Start on node %s (pnid=%d)\n", method_name, __LINE__, Node[Monitor->TmSyncPNid]->GetName(), Monitor->TmSyncPNid);
+ trace_printf("%s@%d - Aborting Slave Sync Start on node %s (pnid=%d)\n", method_name, __LINE__, Node[Monitor->tmSyncPNid_]->GetName(), Monitor->tmSyncPNid_);
MyNode->SetTmSyncState( SyncState_Null );
if (trace_settings & (TRACE_SYNC | TRACE_TMSYNC))
trace_printf("%s@%d - Node %s (pnid=%d) TmSyncState updated (%d)(%s)\n", method_name, __LINE__, MyNode->GetName(), MyPNID, MyNode->GetTmSyncState(), SyncStateString( MyNode->GetTmSyncState() ) );
@@ -2443,7 +2443,7 @@ void CCluster::HandleOtherNodeMsg (struct internal_msg_def *recv_msg,
// Continue with other node's Slave TmSync Start request
if (trace_settings & (TRACE_SYNC | TRACE_TMSYNC))
trace_printf("%s@%d - Slave Sync Start on node %s (pnid=%d)\n", method_name, __LINE__, Node[pnid]->GetName(), pnid);
- TmSyncPNid = pnid;
+ tmSyncPNid_ = pnid;
Node[pnid]->SetTmSyncState( recv_msg->u.sync.state );
if (trace_settings & (TRACE_SYNC | TRACE_TMSYNC))
{
@@ -2467,7 +2467,7 @@ void CCluster::HandleOtherNodeMsg (struct internal_msg_def *recv_msg,
// Continue with other node's Slave TmSync Start request
if (trace_settings & (TRACE_SYNC | TRACE_TMSYNC))
trace_printf("%s@%d - Slave Sync Start on node %s (pnid=%d)\n", method_name, __LINE__, Node[pnid]->GetName(), pnid);
- TmSyncPNid = pnid;
+ tmSyncPNid_ = pnid;
Node[pnid]->SetTmSyncState( recv_msg->u.sync.state );
if (trace_settings & (TRACE_SYNC | TRACE_TMSYNC))
{
@@ -2783,9 +2783,9 @@ void CCluster::HandleMyNodeMsg (struct internal_msg_def *recv_msg,
case SyncType_TmData:
if (trace_settings & (TRACE_SYNC | TRACE_TMSYNC))
trace_printf("%s@%d - TMSYNC(TmData) on Node %s (pnid=%d)\n", method_name, __LINE__, Node[MyPNID]->GetName(), MyPNID);
- TmSyncPNid = MyPNID;
+ tmSyncPNid_ = MyPNID;
if (trace_settings & (TRACE_SYNC | TRACE_TMSYNC))
- trace_printf("%s@%d - Sync communicated, TmSyncPNid=%d\n", method_name, __LINE__, TmSyncPNid);
+ trace_printf("%s@%d - Sync communicated, tmSyncPNid_=%d\n", method_name, __LINE__, tmSyncPNid_);
if ( ! MyNode->IsSpareNode() && MyNode->GetPhase() != Phase_Ready )
{
MyNode->CheckActivationPhase();
@@ -2974,7 +2974,7 @@ void CCluster::InitializeConfigCluster( void )
int rankToPnid[worldSize];
CClusterConfig *clusterConfig = Nodes->GetClusterConfig();
- CurNodes = worldSize;
+ currentNodes_ = worldSize;
if ( IsRealCluster )
{
@@ -3038,10 +3038,10 @@ void CCluster::InitializeConfigCluster( void )
int TmLeaderPNid = -1;
if (IsMaster)
{
- TmLeaderNid = Nodes->GetFirstNid();
- TmLeaderPNid = LNode[TmLeaderNid]->GetNode()->GetPNid();
+ tmLeaderNid_ = Nodes->GetFirstNid();
+ TmLeaderPNid = LNode[tmLeaderNid_]->GetNode()->GetPNid();
}
- // Non-master monitors in AGENT mode in a real cluster initialize all
+ // Monitors processes in AGENT mode in a real cluster initialize all
// remote nodes to a down state. The master monitor and the joining
// monitors will set the joining node state to up as part of the node
// re-integration processing as monitor processes join the cluster
@@ -3179,8 +3179,8 @@ void CCluster::InitializeConfigCluster( void )
delete [] commPortNums;
delete [] syncPortNums;
- TmLeaderNid = Nodes->GetFirstNid();
- int TmLeaderPNid = LNode[TmLeaderNid]->GetNode()->GetPNid();
+ tmLeaderNid_ = Nodes->GetFirstNid();
+ int TmLeaderPNid = LNode[tmLeaderNid_]->GetNode()->GetPNid();
// Any nodes not in the initial MPI_COMM_WORLD are down.
for (int i=0; i<GetConfigPNodesCount(); ++i)
@@ -3220,7 +3220,7 @@ void CCluster::InitializeConfigCluster( void )
}
else
{
- TmLeaderNid = 0;
+ tmLeaderNid_ = 0;
}
// Initialize communicators for point-to-point communications
@@ -3828,7 +3828,7 @@ void CCluster::ReIntegrateMPI( int initProblem )
{ // Already connected to creator monitor
comms_[i] = intraCommCreatorMon;
otherMonRank_[i] = 0;
- ++CurNodes;
+ ++currentNodes_;
// Set bit indicating node is up
upNodes_.upNodes[i/MAX_NODE_BITMASK] |= (1ull << (i%MAX_NODE_BITMASK));
@@ -3922,7 +3922,7 @@ void CCluster::ReIntegrateMPI( int initProblem )
comms_[i] = intraComm;
otherMonRank_[i] = 0;
- ++CurNodes;
+ ++currentNodes_;
Node[i]->SetSyncPort( nodeInfo[i].syncPort );
Node[i]->SetState( State_Up );
@@ -4163,7 +4163,7 @@ void CCluster::ReIntegrateSock( int initProblem )
}
otherMonRank_[nodeInfo[i].pnid] = 0;
- ++CurNodes;
+ ++currentNodes_;
// Store port numbers for the node
strncpy(commPort, nodeInfo[i].commPort, MPI_MAX_PORT_NAME);
@@ -4310,7 +4310,7 @@ void CCluster::ReIntegrateSock( int initProblem )
}
otherMonRank_[nodeInfo[i].pnid] = 0;
- ++CurNodes;
+ ++currentNodes_;
// Store port numbers for the node
strncpy(commPort, nodeInfo[i].commPort, MPI_MAX_PORT_NAME);
@@ -4583,7 +4583,7 @@ void CCluster::setNewComm( int pnid )
close( socks_[pnid] );
socks_[pnid] = -1;
}
- --CurNodes;
+ --currentNodes_;
}
if (trace_settings & TRACE_RECOVERY)
@@ -4595,7 +4595,7 @@ void CCluster::setNewComm( int pnid )
comms_[it->pnid] = it->comm;
otherMonRank_[it->pnid] = it->otherRank;
- ++CurNodes;
+ ++currentNodes_;
// Set bit indicating node is up
upNodes_.upNodes[it->pnid/MAX_NODE_BITMASK] |= (1ull << (it->pnid%MAX_NODE_BITMASK));
@@ -4686,14 +4686,14 @@ void CCluster::setNewSock( int pnid )
shutdown( socks_[pnid], SHUT_RDWR);
close( socks_[pnid] );
socks_[pnid] = -1;
- --CurNodes;
+ --currentNodes_;
}
CNode *node= Nodes->GetNode( it->pnid );
socks_[it->pnid] = it->socket;
sockPorts_[it->pnid] = node->GetSyncSocketPort();
otherMonRank_[it->pnid] = it->otherRank;
- ++CurNodes;
+ ++currentNodes_;
if (trace_settings & (TRACE_INIT | TRACE_RECOVERY))
{
@@ -6040,7 +6040,7 @@ void CCluster::ValidateClusterState( cluster_state_def_t nodestate[],
// Evaluate each active (up) node in the cluster
int pnodesCount = 0;
for (int index = 0;
- index < GetConfigPNodesMax() && pnodesCount < CurNodes;
+ index < GetConfigPNodesMax() && pnodesCount < currentNodes_;
++index)
{
if ( nodestate[index].seq_num != 0 )
@@ -6104,11 +6104,11 @@ void CCluster::ValidateClusterState( cluster_state_def_t nodestate[],
if (trace_settings & (TRACE_SYNC | TRACE_RECOVERY | TRACE_INIT))
{
- trace_printf("%s@%d concurringNodes=%d, CurNodes=%d\n",
- method_name, __LINE__, concurringNodes, CurNodes);
+ trace_printf("%s@%d concurringNodes=%d, currentNodes_=%d\n",
+ method_name, __LINE__, concurringNodes, currentNodes_);
}
- if (concurringNodes == CurNodes)
+ if (concurringNodes == currentNodes_)
{ // General agreement that node is down, proceed to mark it down
CNode *downNode = Nodes->GetNode( it->exitedPnid );
@@ -6149,7 +6149,7 @@ void CCluster::ValidateClusterState( cluster_state_def_t nodestate[],
"%d but only %d of %d nodes also lost the "
"connection. See up: %s. See down: %s. So node "
"%d is going down (at seq #%lld).\n", method_name,
- it->exitedPnid, concurringNodes, CurNodes,
+ it->exitedPnid, concurringNodes, currentNodes_,
setSeesUp.c_str(), setSeesDown.c_str(),
MyPNID, seqNum_ );
mon_log_write(MON_CLUSTER_VALIDATE_STATE_2, SQ_LOG_ERR, buf);
@@ -6198,7 +6198,7 @@ void CCluster::ValidateClusterState( cluster_state_def_t nodestate[],
int pnodesCount2 = 0;
for (int remIndex = 0;
- remIndex < GetConfigPNodesMax() && pnodesCount2 < CurNodes;
+ remIndex < GetConfigPNodesMax() && pnodesCount2 < currentNodes_;
++remIndex)
{
bool someExited = false;
@@ -6248,7 +6248,7 @@ void CCluster::ValidateClusterState( cluster_state_def_t nodestate[],
{ // This remote node sees node pnid as up
int pnodesCount3 = 0;
for (int exitedPNid = 0;
- exitedPNid < GetConfigPNodesMax() && pnodesCount3 < CurNodes;
+ exitedPNid < GetConfigPNodesMax() && pnodesCount3 < currentNodes_;
++exitedPNid)
{
CNode *exitedNode = Nodes->GetNode( /*indexToPnid_[remIndex]*/exitedPNid );
@@ -6666,7 +6666,7 @@ void CCluster::UpdateClusterState( bool &doShutdown,
abort();
}
Node[index]->SetState( State_Down );
- --CurNodes;
+ --currentNodes_;
// Clear bit in set of "up nodes"
upNodes_.upNodes[index/MAX_NODE_BITMASK] &= ~(1ull << (index%MAX_NODE_BITMASK));
}
@@ -6738,7 +6738,7 @@ void CCluster::UpdateClusterState( bool &doShutdown,
// Programmer bonehead!
abort();
}
- --CurNodes;
+ --currentNodes_;
// Clear bit in set of "up nodes"
upNodes_.upNodes[index/MAX_NODE_BITMASK] &= ~(1ull << (index%MAX_NODE_BITMASK));
@@ -7068,14 +7068,14 @@ bool CCluster::checkIfDone ( )
if (trace_settings & TRACE_SYNC_DETAIL)
trace_printf("%s@%d - Node %d shutdown level=%d, state=%s. Process "
- "count=%d, internal state=%d, CurNodes=%d, "
+ "count=%d, internal state=%d, currentNodes_=%d, "
"local process count=%d\n",
method_name, __LINE__, MyNode->GetPNid(),
MyNode->GetShutdownLevel(),
StateString(MyNode->GetState()),
Nodes->ProcessCount(),
MyNode->getInternalState(),
- CurNodes, MyNode->GetNumProcs());
+ currentNodes_, MyNode->GetNumProcs());
// Check if we are also done
if (( MyNode->GetState() != State_Down ) &&
@@ -7094,7 +7094,7 @@ bool CCluster::checkIfDone ( )
return false;
}
else if ( (Nodes->ProcessCount() <=
- (CurNodes*MAX_PRIMITIVES)) // only WDGs alive
+ (currentNodes_*MAX_PRIMITIVES)) // only WDGs alive
&& !MyNode->isInQuiesceState() // post-quiescing will
// expire WDG (cluster)
&& !waitForWatchdogExit_ ) // WDG not yet exiting
http://git-wip-us.apache.org/repos/asf/trafodion/blob/3d7855b6/core/sqf/monitor/linux/cluster.h
----------------------------------------------------------------------
diff --git a/core/sqf/monitor/linux/cluster.h b/core/sqf/monitor/linux/cluster.h
index 6b658ae..ff49e56 100644
--- a/core/sqf/monitor/linux/cluster.h
+++ b/core/sqf/monitor/linux/cluster.h
@@ -124,12 +124,12 @@ public:
void DoDeviceReq(char * ldevname);
void ExpediteDown( void );
- inline int GetTmLeader( void ) { return( TmLeaderNid); }
- inline void SetTmLeader( int tmLeaderNid ) { TmLeaderNid = tmLeaderNid; }
- inline int GetMonitorLeader( void ) { return( MonitorLeaderPNid); }
- inline void SetMonitorLeader( int monitorLeaderPNid ) { MonitorLeaderPNid = monitorLeaderPNid; }
+ inline int GetTmLeader( void ) { return( tmLeaderNid_ ); }
+ inline void SetTmLeader( int tmLeaderNid ) { tmLeaderNid_ = tmLeaderNid; }
+ inline int GetMonitorLeader( void ) { return( monitorLeaderPNid_); }
+ inline void SetMonitorLeader( int monitorLeaderPNid ) { monitorLeaderPNid_ = monitorLeaderPNid; }
int GetDownedNid( void );
- inline int GetTmSyncPNid( void ) { return( TmSyncPNid ); } // Physical Node ID of current TmSync operations master
+ inline int GetTmSyncPNid( void ) { return( tmSyncPNid_ ); } // Physical Node ID of current TmSync operations master
void InitClusterComm(int worldSize, int myRank, int *rankToPnid);
void addNewComm(int nid, int otherRank, MPI_Comm comm);
void addNewSock(int nid, int otherRank, int sockFd );
@@ -210,7 +210,7 @@ protected:
CNode **Node; // array of nodes
CLNode **LNode; // array of logical nodes
- int TmSyncPNid; // Physical Node ID of current TmSync operations master
+ int tmSyncPNid_; // Physical Node ID of current TmSync operations master
void AddTmsyncMsg( struct sync_buffer_def *tmSyncBuffer
@@ -229,15 +229,14 @@ protected:
CLock syncCycle_;
private:
- int CurNodes; // Current # of nodes in the cluster
- int CurProcs; // Current # if processes alive in MPI_COMM_WORLD
+ int currentNodes_; // Current # of nodes in the cluster
int configPNodesCount_; // # of physical nodes configured
int configPNodesMax_; // max # of physical nodes that can be configured
- int *NodeMap; // Mapping of Node ranks to COMM_WORLD ranks
- int TmLeaderNid; // Nid of currently assigned TM Leader node
- int MonitorLeaderPNid; // PNid of currently assigned Monitor leader node
- int tmReadyCount_; // # of DTM processes ready for transactions
- size_t minRecvCount_; // minimum size of receive buffer for allgather
+ int *nodeMap_; // Mapping of Node ranks to COMM_WORLD ranks
+ int tmLeaderNid_; // Nid of currently assigned TM Leader node
+ int monitorLeaderPNid_; // PNid of currently assigned Monitor leader node
+ int tmReadyCount_; // # of DTM processes ready for transactions
+ size_t minRecvCount_; // minimum size of receive buffer for allgather
// Pointer to array of "sync_buffer_def" structures. Used by
// ShareWithPeers in "Allgather" operation.
http://git-wip-us.apache.org/repos/asf/trafodion/blob/3d7855b6/core/sqf/monitor/linux/pnode.cxx
----------------------------------------------------------------------
diff --git a/core/sqf/monitor/linux/pnode.cxx b/core/sqf/monitor/linux/pnode.cxx
index 485d013..783640f 100644
--- a/core/sqf/monitor/linux/pnode.cxx
+++ b/core/sqf/monitor/linux/pnode.cxx
@@ -1709,7 +1709,6 @@ void CNodeContainer::AddNodes( )
else
{
if (pnid >= maxNode) // only for workstation acting as single node
-// || (IsAgentMode &&(strcmp( MasterMonitorName, Node_name ) != 0)))
{
rank = -1; // -1 creates node in down state
}
http://git-wip-us.apache.org/repos/asf/trafodion/blob/3d7855b6/core/sqf/monitor/linux/tmsync.cxx
----------------------------------------------------------------------
diff --git a/core/sqf/monitor/linux/tmsync.cxx b/core/sqf/monitor/linux/tmsync.cxx
index 60d9f40..b87f0f4 100644
--- a/core/sqf/monitor/linux/tmsync.cxx
+++ b/core/sqf/monitor/linux/tmsync.cxx
@@ -321,7 +321,7 @@ int CTmSync_Container::CoordinateTmDataBlock ( struct sync_def *sync )
exchangeTmSyncData( sync, false );
syncCycle_.unlock();
ExchangeTmSyncState( false );
- if (( Monitor->TmSyncPNid == MyPNID ) &&
+ if (( Monitor->tmSyncPNid_ == MyPNID ) &&
( Nodes->GetTmState( SyncState_Start ) == SyncState_Start ) )
{
// send unsolicited messages to other TMs in
@@ -353,7 +353,7 @@ int CTmSync_Container::CoordinateTmDataBlock ( struct sync_def *sync )
else
{
if (trace_settings & (TRACE_SYNC | TRACE_TMSYNC))
- trace_printf("%s@%d" " - Tm Sync failed to start, TmSyncPNid=%d, MyPNID=%d, " "TmSyncState=%d, expecting=%d\n", method_name, __LINE__, TmSyncPNid, MyPNID, Nodes->GetTmState( SyncState_Start ), SyncState_Start);
+ trace_printf("%s@%d" " - Tm Sync failed to start, tmSyncPNid_=%d, MyPNID=%d, " "TmSyncState=%d, expecting=%d\n", method_name, __LINE__, tmSyncPNid_, MyPNID, Nodes->GetTmState( SyncState_Start ), SyncState_Start);
if (MyNode->GetTmSyncState() == SyncState_Start)
{
MyNode->SetTmSyncState( SyncState_Null );
@@ -449,7 +449,7 @@ void CTmSync_Container::EndTmSync( MSGTYPE type )
{
trace_printf("%s@%d - Request (%p) nid=%d, handle=%d, tag=%d, unsol=%d, comp=%d\n", method_name, __LINE__, req, req->Nid, req->Handle, req->Tag, req->Unsolicited, req->Completed);
}
- if ( TmSyncPNid == MyPNID )
+ if ( tmSyncPNid_ == MyPNID )
{
if ( MyNode->GetLNodesCount() > 1 )
{
@@ -666,7 +666,7 @@ void CTmSync_Container::ProcessTmSyncReply( struct message_def * msg )
TmSyncReplyCode |= msg->u.reply.u.unsolicited_tm_sync.return_code;
tmsync_req->Completed = true;
UnsolicitedComplete( msg );
- if ( TmSyncPNid == MyPNID )
+ if ( tmSyncPNid_ == MyPNID )
{
if (trace_settings & (TRACE_REQUEST | TRACE_TMSYNC))
trace_printf("%s@%d - Local Unsolicited TmSync reply, handle="
@@ -1102,7 +1102,7 @@ bool CTmSync_Container::TmSyncPending( void )
trace_printf("%s@%d" " - PendingTmSync=%d, total=%d, replies=%d, pending=%d\n", method_name, __LINE__, PendingSlaveTmSync, GetTotalSlaveTmSyncCount(), GetTmSyncReplies(), GetPendingSlaveTmSyncCount() );
if (( MyNode->GetTmSyncState() == SyncState_Abort ) &&
- ( TmSyncPNid != MyPNID ) &&
+ ( tmSyncPNid_ != MyPNID ) &&
( GetTmSyncReplies() == GetTotalSlaveTmSyncCount() ) )
{
CommitTmDataBlock( MPI_ERR_UNKNOWN );
http://git-wip-us.apache.org/repos/asf/trafodion/blob/3d7855b6/core/sqf/monitor/linux/zclient.cxx
----------------------------------------------------------------------
diff --git a/core/sqf/monitor/linux/zclient.cxx b/core/sqf/monitor/linux/zclient.cxx
index 107cf32..1c133ca 100644
--- a/core/sqf/monitor/linux/zclient.cxx
+++ b/core/sqf/monitor/linux/zclient.cxx
@@ -523,6 +523,8 @@ const char* CZClient::WaitForAndReturnMaster( bool doWait )
{
break;
}
+ usleep(1000000); // sleep for a second as to not overwhelm the system
+ retries++;
continue;
}
else if ( rc == ZOK )
@@ -549,14 +551,14 @@ const char* CZClient::WaitForAndReturnMaster( bool doWait )
break;
}
usleep(1000000); // sleep for a second as to not overwhelm the system
- retries++;
+ retries++;
continue;
}
}
else // error
{
- if (trace_settings & (TRACE_INIT | TRACE_RECOVERY))
+ if (trace_settings & (TRACE_INIT | TRACE_RECOVERY))
{
trace_printf( "%s@%d Error (MasterMonitor) WaitForAndReturnMaster returned rc (%d), retries %d\n"
, method_name, __LINE__, rc, retries );
@@ -946,6 +948,8 @@ int CZClient::CreateMasterZNode( const char *nodeName )
, "[%s], RegisterZNode(%s) failed with error %s\n"
, method_name, monData.c_str(), zerror(rc) );
mon_log_write(MON_ZCLIENT_CREATEMASTERZNODE, SQ_LOG_ERR, buf);
+
+ TRACE_EXIT;
return(rc); // Return the error
}
if (trace_settings & (TRACE_INIT | TRACE_RECOVERY))