OpenStack High Availability (HA)-Mariadb

# Galera


* * Features * *

-Full replication, transactions are either committed at all nodes, or are rolled back
-Multi-master replication for read/write operations at any node
-In-parallel application of events from the server, true parallel replication
-Node Auto-configuration, the fault node is automatically removed from the cluster. Once the failed node joins the cluster again, there is no need to manually back up the current database and copy to the failed node.
-No delay in synchronization to ensure consistency of data
-With application compatibility, no need to change application, with native MySQL interface
-Each node contains a complete copy of the data
-synchronous replication of individual nodes, not implemented through Binlog, but through Galera.cache.

* * In the production environment, it is recommended to configure 3 nodes in the cluster, otherwise it is easy to produce brain fissure. **


* * Architecture Benefits: * *

-A true multi-master architecture, where any node can read/write without the need for a read/write separation.
-No centralized management, can lose any node at any point in time, the normal work of the cluster will not be affected.
-node downtime does not result in data loss.
-Transparent to the app, no need to change the app or just make minimal changes.

* * Architecture Disadvantages * *

-The cost of joining the new node is large and the complete data needs to be copied.
-The write extension problem cannot be resolved effectively, and all write operations will occur on all nodes.
-How many nodes there are to duplicate data.
-Because transactional commits require cross-node communication (distributed transactions), writes are much slower than master-slave replication, and as the cluster nodes increase, writes become slower, and natural deadlocks and rollbacks are more frequent.
-Network requirements are very high, if the network fluctuations or the computer room is ARP attack, so that two nodes are lost, Galera cluster cluster will have a brain fissure, service will not be available.
-The write throughput of the entire cluster is limited by the weakest node, and if one node becomes slow, it will affect the entire cluster being slow.

OpenStack High Availability (HA)-Mariadb