TL;DR: Make sure to run “SET persist_only disabled_storage_engines=’MyISAM’, persist sql_generate_invisible_primary_key=ON;” on all instances and restart each one in your MySQL InnoDB Cluster.

Ok, what does “safe from naughtiness” mean?:
– Anyone creating tables that aren’t InnoDB, as this doesn’t make sense, after all, it is an “InnoDB” cluster.
– Making sure all tables have a Primary Key (invisible or not).
– Making sure that my (invisible) primary keys are visible to the cluster as it will rightfully complain if they aren’t!

This basically means that once you’ve got it all up and running you won’t run into those horrible situations whereby someone, somewhere, creates a MyISAM table that didn’t have a Primary Key and thus leave you with a broken cluster.

Eg.

MySQL rtnode-01:3306 ssl JS > vlc.status()
{
 "clusterName": "VLC",
 "clusterRole": "PRIMARY", …

Now this is far from being any observability manual for your InnoDB Cluster and let alone go into everything MySQL Shell API Admin, or the collectDiagnostics utility. You can also use the default javascript commands that we all know and love via dba.getCluster() and so on, but here’s a different take.

I just want to share something I’ve been playing with to pull out some key info from mycluster. Hope it helps someone else out there.

General setup:

select cluster_id, cluster_name, description, cluster_type, primary_mode, clusterset_id
from mysql_innodb_cluster_metadata.clusters;

Members of our cluster:

select * from …

Back again, now with MySQL installs. And this means using the MySQL repository this time around.

I’ve been installing and configuring InnoDB Clusters and ClusterSets and thinking about the Ansible and Terraform users amongst us, maybe one-liners might help someone out there.

So, what about if I share how to install the MySQL repo, install the MySQL instance, create an InnoDB Cluster, add a MySQL Router, create a ClusterSet, make sure the Router is ClusterSet-aware, and then test it out. And all via one-liners.

First up, obrigado Miguel for https://github.com/miguelaraujo/ClusterSet-Demo.

To simplify the command execution sequence, these sections aim to help summarize the technical commands required to create the whole platform. And on a default path & port configuration, to ease operational deployments for all those 000’s of …

In the first part of this short series, we went through the build of a full Clusterset architecture, which included 2 full clusters (3 members each) and an additional cluster with one member, which can serve different purposes. With the MySQL ClusterSet, there are new capabilities with the router instances. These features enable valuable options for users of… Read More »

InnoDB Cluster has been around for what feels like a long time. It is the core platform for MySQL High Availability. InnoDB Cluster NOW extends that core feature into a platform that also enables DR support where multiple Disaster Recovery Regions are capable.

The power of PlanetScale within a beautiful GUI — now you can connect to development branches switch between them and even create deploy requests in Arctype.

We already saw how easy it’s to deploy solutions on OCI using Terraform and Resource Manager’s Stack. I’ve published several resources available on this page.

Today we will see how easy it’s to deploy Apache Superset on OCI using MySQL Data Service.

Apache Superset is an open source BI, Reporting, Charting tool that competes with Tableau, Looker, etc.  For a list of companies that have deployed Superset, see:  https://github.com/apache/superset/blob/master/RESOURCES/INTHEWILD.md.

Superset is loaded with options that make it easy for users of all skill sets to explore and visualize their data, from simple line charts to highly detailed geospatial charts.

For a gallery of …

We’re excited to share that Marcelo Altmann from the Percona Server Engineering Team has just been promoted from Oracle ACE Associate to Oracle ACE.

Congratulations!

The Oracle ACE Program recognizes and rewards community members for their technical contributions to the Oracle community.

Marcelo initially joined Percona as a senior support engineer in our global services organization, where he helped customers with running their MySQL-based environments. In early 2020, he joined our Server Engineering team and has been actively involved in the development of Percona Server for MySQL and Percona …

In part 1 of the series about Point-in-Time Recovery in OCI MDS, we saw how to stream the binary log to Object Storage.

In this blog post, we will see how we can restore the data up to certain point.

The high level process is simple:

1. we need to restore the last backup just before the point we want to return to
2. find the last GTID that was executed in that backup (not 100% mandatory but can accelerate the process)
3. choose what is the objective:
   1. just skip one transaction (following transaction could depend on it)
   2. recover up to that transaction and skip all the following ones
4. replay the binary logs taking in consideration the chosen option of point 3

The scenario …

To setup point-in-time recovery for MDS using Object Storage, these are the prerequisites:

• a MDS instance running
• a backup plan (default)
• a compute instance
• an Object Storage Bucket

In this article, I won’t focus on how to create a MDS instance, a compute instance, enable backups and Object Storage Bucket, this is easy and there is already a lot of literature about them (you can also check several of my previous articles).

This is part 1 of the Point-in-Time Recovery in OCI MDS series. It’s only about streaming the binary logs to Object Store. In part 2, we will see how to use them and perform PITR.

Saving the binary logs

In fact most of the work will be to setup the compute instance to store the binary logs coming from MDS (one single compute instance could stream those binlogs from multiple MDS source).

On the compute instance we need …