As i understand Single message is sent when new node joins cluster , or when node dynamically creates new cache. So it sends Single message to coordinator and receives GridDhtPartitionsFullMessage(holding edited partitions info).
Still hardly can understand how could it be two GridDhtPartitionsSingleMessage with the same exchangeId? What is the scenario of this situation чт, 4 мая 2017 г. в 14:45, ALEKSEY KUZNETSOV <alkuznetsov...@gmail.com>: > Thank you! > > чт, 4 мая 2017 г. в 12:25, Dmitry Pavlov <dpavlov....@gmail.com>: > >> Thank you for clarification. >> >> All Single Message will have same exchange ID because single messages are >> issued in one round of ExchangeFuture (one complete round of exchange is >> to >> be finished before next will start). >> >> >> Exchange ID on non coordinator nodes is filled in single messages >> (GridDhtPartitionsSingleMessage) from >> GridDhtPartitionsExchangeFuture#exchangeId field. Value of exchange ID is >> the same in all nodes and content of exchange future is determined by >> original event caused this exchange (e.g. node left, node joined, node >> failed) for current topology version. >> >> >> чт, 4 мая 2017 г. в 11:32, ALEKSEY KUZNETSOV <alkuznetsov...@gmail.com>: >> >> > I know this. >> > But the question was as follows. >> > Imagine, remaining set contains 2 nodes. It implies that both nodes had >> > sent two GridDhtPartitionsSingleMessage with the same exchangeId. >> > So, how could it possible 2 messages from different nodes has got equal >> > exchangeId. >> > >> > ср, 3 мая 2017 г. в 19:23, Dmitry Pavlov <dpavlov....@gmail.com>: >> > >> > > Hi, Aleksey, >> > > >> > > empty remaining set >> > > >> > > >> > >> (org.apache.ignite.internal.processors.cache.distributed.dht.preloader.GridDhtPartitionsExchangeFuture#remaining) >> > > is the mandatory condition of sendAllPartitions() to be activated. >> > > >> > > When coordinator receives single partition message ( >> > > GridDhtPartitionsSingleMessage), it removes one node from remaining >> set. >> > > The same is true for node left message. >> > > >> > > Best Regards, >> > > Dmitry Pavlov >> > > >> > > ср, 3 мая 2017 г. в 19:10, ALEKSEY KUZNETSOV < >> alkuznetsov...@gmail.com>: >> > > >> > > > Hi, Igntrs! When processing single partitions message in >> > > > GridDhtPartitionsExchangeFuture#processMessage() then , coordinator >> > could >> > > > answer with GridDhtPartitionsFullMessage >> > > > from sendAllPartitions. >> > > > But the field GridDhtPartitionsExchangeFuture#remaining must be >> empty >> > in >> > > > this case. >> > > > >> > > > I wounder, how could it be empty in case if "remaining" held 2 >> nodes, >> > or >> > > > more ? >> > > > >> > > > >> > > > -- >> > > > >> > > > *Best Regards,* >> > > > >> > > > *Kuznetsov Aleksey* >> > > > >> > > >> > -- >> > >> > *Best Regards,* >> > >> > *Kuznetsov Aleksey* >> > >> > -- > > *Best Regards,* > > *Kuznetsov Aleksey* > -- *Best Regards,* *Kuznetsov Aleksey*
