Everywhere I go to present MySQL InnoDB Cluster, during the demo of creating a cluster, many people doesn’t understand why when I’ve 2 members, my cluster is not yet tolerant to any failure.

Indeed when you create a MySQL InnoDB Cluster, as soon as you have added your second instance, you can see in the status:

    "status": "OK_NO_TOLERANCE",      
    "statusText": "Cluster is NOT tolerant to any failures.",

Quorum

Why is that ? It’s because, to be part of primary partition (the partition that holds the service, the one having a Primary-Master in Single Primary Mode, the default mode), your partition must reach the majority of nodes (quorum). In MySQL InnoDB Cluster (and many other cluster solutions), to achieve quorum, the amount of members in a partition must be > (bigger) than 50%.

So when we have 2 nodes, if there is a network issue between the …

Consistency during reads have been a small concern from the adopters of MySQL InnoDB Cluster (see this post and this one).

This is why MySQL supports now (since 8.0.14) a new consistency model to avoid such situation when needed.

Nuno Carvalho and Aníbal Pinto already posted a blog series I highly encourage you to read:

• Group Replication – Consistency Levels
• Group Replication: Preventing stale reads on …

MySQL InnoDB Cluster is evolving very nicely. I realized that the MySQL Shell also improved a lot and that it has never been so easy to setup a cluster on 3 new nodes.

This is a video of the updated procedure on how to install MySQL InnoDB Cluster on GNU Linux rpm based (Oracle Linux, RedHat, CentOS, Fedora, …)

This event was just awesome. We got 110 participants ! Thank you everybody and also a big thank to the speakers.

Here are the slides of all the sessions:

• MySQL Shell – A DevOps-engineer day with MySQL’s development and administration tool, Miguel Araújo
• MySQL Shell : the best DBA tool ? – How to use the MySQL Shell as a framework for DBAs, Frédéric Descamps
• Tuning MySQL 8.0 InnoDB for High Load, Dimitri Kravtchuk
• …

Recently we had a customer who came to us for help with monitoring their InnoDB cluster setup. They run a 5.7 InnoDB cluster and suffered from a network outage that split up their cluster leaving it with no writeable primaries. As part of their incident followup, they asked us for methods to monitor this cluster.

I had tested before with InnoDB clusters (in both 8.0 and 5.7 variants) and I was confident that we could parse the cluster node “role” (read-write aka primary vs. read-only aka secondary) from the performance_schema tables. As it turned out, this feature is not in 5.7 but only 8.0. However, the docs on this are wrong for 5.7 as these docs suggest that the performance_schema.group_replication_members table would show PRIMARY and SECONDARY role of each cluster node. I have submitted a …

MySQL InnoDB is more and more popular. The adoption of it is even faster than I expected. Recently, during my travel in Stockholm, Sweden, a customer asked me what was the best practice to backup a cluster.

Since my interlocutor was a customer, the obvious choice is to use MySQL Enterprise Backup (known as MEB). Of course any other physical backup should be also fine.

The customer told me he was using cron to schedule his backup and was only using full backups… That’s perfect. So I told him that there is nothing complicated and that the cron job should something like:

mysqlbackup --with-timestamp --backup-dir /backup  backup

Of course, I do not recommend the use of --user clusteradmin --password=xxxxxin the cronjob but configure your crendentials using …

For the third year in a row, we will take advantage of the mass presence of our MySQL Engineers during FOSDEM to organize the pre-FOSDEM MySQL Day.

The program of this 3rd edition is already on track, thank you to all the speakers who already confirmed their participation.

In the first part of this howto, I illustrated how to setup two MySQL InnoDB Cluster linked by an asynchronous replication.

In that solution, I didn’t use any replication filters to ignore the replication of the InnoDB Cluster’s metadata (mysql_innodb_cluster_metadata), but I used the same metadata tables with two different clusters in it.

The benefit is that this allows to backup everything from any node in any of the data center, it works also in MySQL 5.7, and there is not risk to mess up with the replication filters.

In this blog I will show how to use replication filters to link two different clusters. This doesn’t work on …

As you know, MySQL InnoDB Cluster is a High Availability solution for MySQL. However more and more people are trying to use it as a Disaster Recovery solution with 2 data centers. Natively, this is not yet supported. But it’s already possible to realize such setup if we agree with the following points:

•  a human interaction is required in case of Disaster Recovery which, by my own experience, is often acceptable
• a human interaction is required if the any Primary-Master acting as asynchronous slave leaves its group  (crash, network problem, …) or becomes secondary

These are not big constraints and it’s relatively easily to deal with them.

The Architecture

The situation is as follow:

• 2 data centers (one active, one inactive, only used for disaster recovery)
• 2 MySQL InnoDB Clusters (one in each DC)
• 3 members in each …

Last time I mentioned four great MySQL books for 2018.  I was tactfully reminded of two books I overlooked. First is Dr. Charles Bell's Introducing InnoDB Cluster which I have not read (but it is on order).

And last, but not least, is Mikael Ronstrum's MySQL Cluster 7.5 Inside and Out.  This is another book on NDB cluster and is a 'msut have' for those running NDB clusters.