Project: MySQL clustering

[tamalsaha]

• [ ] https://githubengineering.com/mitigating-replication-lag-and-reducing-read-load-with-freno/
• [x] Group replication single-primary - v0.12.0 (MySQL 5.7.25)
• [ ] Group replication multi-primary - (MySQL 5.7.25)
• [ ] Group replication for MySQL 8.x
• [ ] Percona XtraDB - v0.13.0
• [ ] MySQL Galera Cluster
• [ ] MariaDB Galera Cluster
• [ ] https://github.com/github/orchestrator
• [ ] Group replication with InnoDB

[tamalsaha]

xref: https://github.com/kubedb/project/issues/79

[ConnorBarnhill]

@tamalsaha Is there any plan to add support for proxy services? IE when creating the cluster create 1.) a proxy that points to the primary for writes and changes when auto-failover occurs, and 2.) a load balancer that distributes reads across either a.) read replicas only or b.) all pods? Not having to reconfigure applications to point to a new host in the event of a primary failure would reduce downtime.

Thanks!

[dan-j]

@tamalsaha I second @ConnorBarnhill's comment!

Would love to see some proxy service like <mysql-name>-master and <mysql-name>-slave which has their endpoints automatically updated when a failover occurs.

Thoughts?

[tamalsaha]

@shudipta is working on ProxySQL support. We are going to look into that after it is ready for testing.

[xuhaowang]

@tamalsaha Will you use the mysql-router to support the InnoDB cluster in the future?

[weenxin]

hello @tamalsaha @shudipta , Oracle MySQL Operator have a different way to implement MySQL InnoDB Cluster.

1. A MySQL InnoDB Cluster contains a Statefulset And a Service
2. Any pod in Statefulset contains two containers: Mysqld and MySQL Agent
3. A Service which backend is the master nodes

MySQL Agent is very interesting, It will monitoring if the MySQLd in the same pod status, If something went wrong , it Will :

1. Readness Probe fail
2. Rejoin to cluster
3. Change Labels to master/secondary for master service

I think kubedb has a very good abstraction(Database Concept/Schema), But kubedb provisioned a mysql cluster, leaving the cluster to Kubernetes, And That is not a HA Solution. Oracle MySQL Operator has been inactive for a long time, If we can Apply/Implement MySQL Agent for KubeDB it will be very Great.

I Want to do this job. However, I don't know if I missed something, Any suggestions ? @tamalsaha @shudipta

[jmordica]

Hi there, an progress on multi-primary?