In the first part of this series, we deployed a 3-node MariaDB Galera Cluster on Ubuntu 24.04.
While a 3-node topology provides the best fault tolerance, sometimes you need a simpler setup – for example, a two-node cluster with a lightweight arbitrator to maintain quorum without running a full third MariaDB instance.
At Vettabase, we often use this pattern in small or resource-limited environments, where running three full database nodes would be overkill.
In this guide, we’ll convert the third database node (Galera3) into a Garbd (Galera Arbitrator Daemon) instance – a tiny yet essential component that helps maintain cluster quorum efficiently.
Installing garbd
sudo apt update sudo apt install galera-arbitrator-4 -y
Configure garbd
Create /etc/default/garb
GALERA_NODES="172.31.2.197:4567,172.31.3.237:4567" GALERA_GROUP="vettabase_galera"
Stop MariaDB and Start garbd
sudo systemctl stop mariadb sudo systemctl start garb
Verify garbd’ Cluster Membership
Once you start the arbitrator service, check its status:
sudo systemctl status garb
You should see output similar to:
INFO: 1.0 (galera1): State transfer to 0.0 (garb) complete. INFO: Member 1.0 (galera1) synced with group.
This confirms that garb has successfully joined the cluster and synchronized with the group.
Check from a Galera Node
On either database node, verify that the cluster size increased to 3:
mariadb -u root -p -S /run/mysqld/mysqld.sock \ -e "SHOW GLOBAL STATUS LIKE 'wsrep_cluster_size';"
Expected output:
+--------------------+-------+ | Variable_name | Value | +--------------------+-------+ | wsrep_cluster_size | 3 | +--------------------+-------+
You can also check the cluster view in the logs – it should show garb alongside your data nodes:
[Note] WSREP: ================================================
View:
id:8bb99c56-be58-11f0-b57c-7e793e9a515a:518
status:primary
protocol_version:4
capabilities:MULTI-MASTER,CERTIFICATION,PARALLEL_APPLYING,REPLAY,ISOLATION,PAUSE,CAUSAL_READ,INCREMENTAL_WS,UNORDERED,PREORDERED,STREAMING,NBO
final:no
own_index:1
members(3):
0:370d8f1e-be6c-11f0-81ee-6ad9baf05d85,garb
1:52a0c379-be67-11f0-b8ae-3201eb641669,galera1
2:c66106d9-be64-11f0-a06a-33bd45771681,galera2
=================================================
The arbitrator node (garb) is now part of the Galera group and maintains quorum without storing any data or running a full MariaDB instance.
Summary
The Galera cluster now runs with two full data nodes and one lightweight arbitrator. The garbd daemon ensures quorum and availability while consuming minimal resources – ideal for cost-sensitive or resource-constrained environments.
Mykhaylo Rykmas
Vettabase is a MariaDB Foundation sponsor!
0 Comments