1. Glossary/

wsrep_cluster_size

Galera cluster size status variable

wsrep_cluster_size is a status variable exposed by the wsrep (Galera) plugin that reports how many nodes are currently connected and synchronized in the cluster. It is not a configuration parameter: its value is dynamic and changes in real time as the cluster topology shifts.

How it works #

Read the value with a simple query:

SHOW STATUS LIKE 'wsrep_cluster_size';

In a correctly operating 3-node cluster the expected result is 3. Galera calculates quorum using the formula ⌊N/2⌋ + 1: with 3 nodes the minimum quorum is 2. If wsrep_cluster_size drops to 1, the surviving node cannot reach quorum, sets wsrep_cluster_status to non-Primary, and rejects writes to prevent split-brain.

Operational context #

Checking wsrep_cluster_size is the first step when an application reports write errors on a Galera cluster. A value lower than the expected node count does not necessarily signal a critical outage: two out of three nodes still hold quorum and the cluster remains in Primary state. The problem arises only when the majority is lost.

In 2-node clusters (a configuration discouraged in production) any single node loss brings wsrep_cluster_size to 1 and immediately blocks writes, requiring either a manual bootstrap or an arbitrator node (garbd).

Notes #

wsrep_cluster_size reflects the size of the current cluster view (Primary Component), not the total number of configured nodes. A node in Donor/Desynced or Joiner state may temporarily not appear in the count, or may lower it, during an SST/IST transfer.