4 ms·
Why not InnoDB Cluster[1] with single primary? There is a slight decrease in write throughput as it is effectively similar to a synchronous replication on a maj
by deanmoriarty 8y ago
Why not InnoDB Cluster[1] with single primary? There is a slight decrease in write throughput as it is effectively similar to a synchronous replication on a majority, but you get native failover and very easy cluster membership management.
[1] https://dev.mysql.com/doc/refman/8.0/en/mysql-innodb-cluster-userguide.html https://dev.mysql.com/doc/refman/8.0/en/mysql-innodb-cluster...
- deanmoriarty 8y agoWould really love to know why this got downvoted instead of replying "InnoDB Cluster is not good because..."? It's actually the most recommended way to do replication from MySQL now, they are heavily deprecating everything around traditional replication, they even removed MySQL Fabric (the proxy for the client to work well with failover) in favor of MySQL Router, which works with just InnoDB cluster. HN, go figure...
- philaroo 8y agoSmall clarification: MySQL Router usage does not require InnoDB cluster; although agreed that most Router development is focused there.
- gtowey 8y agoThat's only practical for a single datacenter. You still really need asyc replication across geographic boundaries with a failover mechanism.
- deanmoriarty 8y agoThank you for the interesting take, it makes sense. Do you think that if you were to be deploying just in multiple availability zones (a-la AWS, so extremely low network latency but good physical/logical isolation) the same concerns would still apply? Thanks