Hi, According javadoc: "Node will stop if and only if it does not store any unique partitions, that don't have another copies in the cluster." [1].
If you want to move data from nodes, you have to delete it from baseline manually or by means of baseline auto adjustment [2]. 1. https://ignite.apache.org/releases/latest/javadoc/org/apache/ignite/ShutdownPolicy.html#GRACEFUL 2. https://ignite.apache.org/docs/latest/starting-nodes#shutting-down-nodes пн, 9 дек. 2024 г. в 13:17, <jrov...@identy.io>: > > Hi. > > > > We are working with Ignite 2.16 and we found some… issues? > > > > We configure Ignite with the following options: > setShutdownPolicy(ShutdownPolicy.GRACEFUL), persistent cache, > CacheAtomicityMode.TRANSACTIONAL, CacheMode.PARTITIONED, > PartitionLossPolicy.READ_ONLY_SAFE, CacheWriteSynchronizationMode.FULL_SYNC, > CacheRebalanceMode.SYNC and 1 backup. > > > > add 6 nodes > activate the cluster > add some data > remove 4 nodes > > > > We expected that the 4 closing nodes sent their data to the remaining 2 > nodes, but some nodes did not close properly (they get blocked without > closing completely), and the data is lost. How does ShutdownPolicy.GRACEFUL > work? > > If we do not use that feature, the data is lost but the nodes close properly > without blocking. > > > > With 2 nodes we had the same problem when we try to remove 1 node. > > > > Thank. > >
