Hi again, >1. If use thick driver, can we use JDBC connection pooling? no, you should not use connection pools in this case. ignite client which is used by the thick driver is concurrent save and can be used by multiple threads and communicate with any server from the cluster.
>2. How/What is the process to backup the data in Apache Ignite? what do you mean by backups? you can set backups number to cache configuration and cluster will store extra copies of you data, so you will be protected from node failure https://apacheignite.readme.io/docs/primary-and-backup-copies to protect your data from cluster failure you can use disk persistance: https://apacheignite.readme.io/docs/distributed-persistent-store there's 3rd party commercial solution for creating snapshots which can be saved on some ntf dir. Also, you propagate data from a cluster to other databases: https://apacheignite.readme.io/docs/3rd-party-store >3. Let's say initially I start Apache Ignite with only one node (cache mode >= replicated, native persistence is turned on). And start populate data into >the cache/table. Now my server node has few millions data. Then I started >another new node in another machine. The data will automatically replicated >to this new node. What will happen if I have a client that connect to this >new node and the replication of data is not yet finish? if you configure the only second server in client's discovery configuration, then the client will connect to the second server, but the client can communicate directly with all server in your cluster, so even if there's still no data on the second server, the client can read data from the first server. ignite is the strongly consistent system, your client will read data from the first server and only when all data are copied to the second server, the client will be able to read them from the second server. The real mechanism is more complicate because ignite will split data by partitions and move partitions to a new server and when a partition is migrated ignite will allow it to be read by client, but any way client always can read only consistent data, there's no way that client can read only the half of data and miss another. -- Sent from: http://apache-ignite-users.70518.x6.nabble.com/
