Bug#29621 - ndb_mgmd can timeout logevent request, making it to fail

parent 2b489b89
...@@ -119,7 +119,11 @@ operator<<(NdbOut& out, const LogLevel & ll) ...@@ -119,7 +119,11 @@ operator<<(NdbOut& out, const LogLevel & ll)
void void
MgmtSrvr::logLevelThreadRun() MgmtSrvr::logLevelThreadRun()
{ {
while (!_isStopThread) { while (!_isStopThread)
{
Vector<NodeId> failed_started_nodes;
Vector<EventSubscribeReq> failed_log_level_requests;
/** /**
* Handle started nodes * Handle started nodes
*/ */
...@@ -145,13 +149,14 @@ MgmtSrvr::logLevelThreadRun() ...@@ -145,13 +149,14 @@ MgmtSrvr::logLevelThreadRun()
if (setEventReportingLevelImpl(node, req)) if (setEventReportingLevelImpl(node, req))
{ {
ndbout_c("setEventReportingLevelImpl(%d): failed", node); failed_started_nodes.push_back(node);
} }
else
{
SetLogLevelOrd ord; SetLogLevelOrd ord;
ord = m_nodeLogLevel[node]; ord = m_nodeLogLevel[node];
setNodeLogLevelImpl(node, ord); setNodeLogLevelImpl(node, ord);
}
m_started_nodes.lock(); m_started_nodes.lock();
} }
} }
...@@ -169,14 +174,17 @@ MgmtSrvr::logLevelThreadRun() ...@@ -169,14 +174,17 @@ MgmtSrvr::logLevelThreadRun()
req.blockRef = _ownReference; req.blockRef = _ownReference;
if (setEventReportingLevelImpl(0, req)) if (setEventReportingLevelImpl(0, req))
{ {
ndbout_c("setEventReportingLevelImpl: failed 2!"); failed_log_level_requests.push_back(req);
} }
} }
else else
{ {
SetLogLevelOrd ord; SetLogLevelOrd ord;
ord = req; ord = req;
setNodeLogLevelImpl(req.blockRef, ord); if (setNodeLogLevelImpl(req.blockRef, ord))
{
failed_log_level_requests.push_back(req);
}
} }
m_log_level_requests.lock(); m_log_level_requests.lock();
} }
...@@ -185,7 +193,28 @@ MgmtSrvr::logLevelThreadRun() ...@@ -185,7 +193,28 @@ MgmtSrvr::logLevelThreadRun()
if(!ERROR_INSERTED(10000)) if(!ERROR_INSERTED(10000))
m_event_listner.check_listeners(); m_event_listner.check_listeners();
NdbSleep_MilliSleep(_logLevelThreadSleep); Uint32 sleeptime = _logLevelThreadSleep;
if (failed_started_nodes.size())
{
m_started_nodes.lock();
for (Uint32 i = 0; i<failed_started_nodes.size(); i++)
m_started_nodes.push_back(failed_started_nodes[i], false);
m_started_nodes.unlock();
failed_started_nodes.clear();
sleeptime = 100;
}
if (failed_log_level_requests.size())
{
m_log_level_requests.lock();
for (Uint32 i = 0; i<failed_log_level_requests.size(); i++)
m_log_level_requests.push_back(failed_log_level_requests[i], false);
m_log_level_requests.unlock();
failed_log_level_requests.clear();
sleeptime = 100;
}
NdbSleep_MilliSleep(sleeptime);
} }
} }
...@@ -1535,7 +1564,6 @@ MgmtSrvr::setEventReportingLevelImpl(int nodeId_arg, ...@@ -1535,7 +1564,6 @@ MgmtSrvr::setEventReportingLevelImpl(int nodeId_arg,
{ {
SignalSender ss(theFacade); SignalSender ss(theFacade);
NdbNodeBitmask nodes; NdbNodeBitmask nodes;
int retries = 30;
nodes.clear(); nodes.clear();
while (1) while (1)
{ {
...@@ -1572,19 +1600,9 @@ MgmtSrvr::setEventReportingLevelImpl(int nodeId_arg, ...@@ -1572,19 +1600,9 @@ MgmtSrvr::setEventReportingLevelImpl(int nodeId_arg,
continue; continue;
} }
// api_reg_conf not recevied yet, need to retry // api_reg_conf not recevied yet, need to retry
break;
}
}
if (nodeId <= max)
{
if (--retries)
{
ss.unlock();
NdbSleep_MilliSleep(100);
continue;
}
return SEND_OR_RECEIVE_FAILED; return SEND_OR_RECEIVE_FAILED;
} }
}
if (nodeId_arg == 0) if (nodeId_arg == 0)
{ {
...@@ -1607,6 +1625,10 @@ MgmtSrvr::setEventReportingLevelImpl(int nodeId_arg, ...@@ -1607,6 +1625,10 @@ MgmtSrvr::setEventReportingLevelImpl(int nodeId_arg,
continue; // node is not connected, skip continue; // node is not connected, skip
if (ss.sendSignal(nodeId, &ssig) == SEND_OK) if (ss.sendSignal(nodeId, &ssig) == SEND_OK)
nodes.set(nodeId); nodes.set(nodeId);
else if (max == nodeId)
{
return SEND_OR_RECEIVE_FAILED;
}
} }
break; break;
} }
...@@ -2988,8 +3010,8 @@ int MgmtSrvr::connect_to_self(void) ...@@ -2988,8 +3010,8 @@ int MgmtSrvr::connect_to_self(void)
return 0; return 0;
} }
template class Vector<unsigned short>;
template class MutexVector<unsigned short>; template class MutexVector<unsigned short>;
template class MutexVector<Ndb_mgmd_event_service::Event_listener>; template class MutexVector<Ndb_mgmd_event_service::Event_listener>;
template class Vector<EventSubscribeReq>;
template class MutexVector<EventSubscribeReq>; template class MutexVector<EventSubscribeReq>;
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment