Load Balancer

Table of contents

• Overview
• Preparations
• Activate MaxScale configuration synchronisation
• Change MaxScale parameters
• Add new slave and make MaxScale known
• Remove old slave and make MaxScale known
• How is the configuration synchronised?
• What happens in the event of a conflict?
• Tests
• Deactivating MaxScale configuration synchronisation again
• Literature/sources

Overview

A feature that I recently discovered while browsing is the MaxScale configuration synchronisation functionality.

This is not primarily about a …

Table of contents

• Overview
• Preparation of the shards (MariaDB database instances)
  • Create test data
  • Creating roles and users
  • MaxScale Monitor User
  • MaxScale Admin User
  • Create application role and accounts
  • Proxy protocol
• MaxScale SchemaRouter configuration
• Starting and stopping the MaxScale Load Balancer
• Application tests
  • Simple application tests
  • New command show shards
  • More general test
  • Less simple tests
  • More complex application tests
  • Cross-shard tests
• Operation of a MaxScale sharding system …

Currently we are in the testing phase of the next release of our Ops Center (focmm). This testing is quite complex because it includes testing the interaction of various different components like a MariaDB Galera Cluster, a Galera Load Balancer, a virtual IP, etc.

Recently I was in the Linuxhotel for a Galera Cluster Training and there one of the other trainers was romanticising about Linux Containers and Incus. So I had a short look, if Incus could be useful to simplify our unit testing. Some old knowledge …

For this how-to we were working with a MariaDB 10.6.5 database server and the newest MariaDB MaxScale 6.3.0 (aka 2.6) load balancer/proxy.

As a starting point we had a working MariaDB Master/Master replication with one active Master (M_a) and one passive Master (M_p). Whereas the passive Master was set to read_only. As a replication user we used the user replication.

Creating database accounts for the MaxScale Load Balancer

The MaxScale load balancer connects itself with the application users to the database. …

Recently we had the case that a customer claimed that MariaDB 10.3 Binary Log is using 150% more space on disk than MySQL 5.7 Binary Log. Because I never observed something similar, but to be honest, I did not look to intensively for this situation, we had to do some clarifications.

First we checked the usual variables which could be candidates for such a behaviour:

binlog_format                = ROW
binlog_row_image             = FULL
binlog_rows_query_log_events = OFF   # MySQL only …

Table of Contents

• Global Options
• Service Options
• Server Options
• Routing Modules
• Listener Options
• Monitor Options
• Filter Options

Global Options

There are some pretty sophisticated Load Balancers for Galera Clusters setups out in the market (ProxySQL, MaxScale, HAproxy, …). They have many different exotic features. You can nearly do everything with them. But this comes at the cost of complexity. Non of them is simple any more.

A widely underestimated Load Balancer solution for Galera Cluster setups is the Galera Load Balancer from Codership. It is an simple Load Balancer solution which serves all of our daily needs when it comes to Galera …

After properly installing and testing a Galera Cluster we see that the set-up is not finished yet. It needs something in front of the Galera Cluster that balances the load over all nodes.
So we install a load balancer in front of the Galera Cluster. Typically nowadays HAProxy is chosen for this purpose. But then we find, that the whole Galera Cluster is still not high available in case the load balancer fails or dies. So we need a second load balancer for high availability.
But how should we properly …

About

glbd is a simple TCP connection balancer made with scalability and performance in mind. It was inspired by pen, but unlike pen its functionality is limited only to balancing generic TCP connections.

Features:

• list of backend servers is configurable in runtime.
• supports server draining, i.e. does not allocate new connections to server, but does not kill existing ones, waiting for them to end gracefully.
• can use epoll API provided by Linux version 2.6 and higher for ultimate routing performance.
• glb is …

Recently we had to set-up a 3-node Galera Cluster with a Load Balancer in front of it. Because Galera Cluster nodes (mysqld) still reply to TCP requests on port 3306 when they are expelled from the Cluster it is not sufficient to just leave it to the Load Balancer to check the port if a Galera node is properly running or not.

We used the wsrep_notify_cmd variable to hook our own script into the Galera Cluster which disables each Node on the Load Balancer when its state changed.

# my.cnf
#
[mysqld] …

Table of Contents

• Building the Galera Load Balancer
• Starting the Galera Load Balancer
• Querying the Galera Load Balancer
• Draining nodes with Galera Load Balancer
• Removing nodes from and adding nodes to Galera Load Balancer

When you install a Galera Cluster for MySQL for High Availability (HA) it is not enough to install the Database Cluster to achieve this goal. You also have to make the application aware of this HA functionality. This is typically done with some kind of load balancing mechanism between …