• unknown's avatar
    Bug#26293 cluster mgmt node sometimes doesn't receive events from all nodes on restart · 9ed1b843
    unknown authored
    - signals where sometimes sent too early when setting up subscriptions
    
    
    ndb/include/kernel/signaldata/DumpStateOrd.hpp:
      added dump for active subscriptions in cmvmi
    ndb/src/kernel/blocks/cmvmi/Cmvmi.cpp:
      added dump for active subscriptions in cmvmi
    ndb/src/mgmsrv/MgmtSrvr.cpp:
      bug in that signals where sent prior to api reg conf arrived, causing thrown away signals and subsequent hangs in mgmtserver
      also add retry if node connected but not yet received api reg conf
    ndb/src/ndbapi/ClusterMgr.cpp:
      added status variable m_api_reg_conf in cluster manager to correctly be able to determine if a node is sendable
    ndb/src/ndbapi/ClusterMgr.hpp:
      added status variable m_api_reg_conf in cluster manager to correctly be able to determine if a node is sendable
    ndb/src/ndbapi/SignalSender.cpp:
      assert to see that node is sendable when signal is sent
    ndb/src/ndbapi/SignalSender.hpp:
      manke metchd const
    9ed1b843
SignalSender.hpp 2.33 KB