Monitoring Maria Cluster

eG Enterprise provides an exclusive Maria Cluster monitoring model that runs quick health checks on the MySQL Cluster at configured intervals, and proactively alerts administrators to potential bottlenecks to the performance of the server.

Figure 1 : The layer model of the MySQL Cluster Component

Using the model depicted by Figure 1, administrators can determine the following:

• Is the cluster primary node active and stable, or has its role changed recently?

• How many primary, secondary, and non-primary nodes are currently connected to the cluster?

• Are there any disconnected, unjoined, or waiting nodes reducing overall cluster reliability?

• Is replication lag time within acceptable limits, or is it abnormally high on any node?

• Are flow control signals (sent or received) occurring too frequently, indicating possible overload or congestion?

• Are too many transactions experiencing errors, or is transaction throughput consistent across nodes?

• Are all nodes synced properly, or are some still in initializing/joining states?

• Are donor nodes present and successfully serving state transfers?

• Are clients able to connect to the cluster nodes, or are there too many connection failures or aborts?

• Is the maximum number of simultaneous connections approaching the configured connection limit?

• Is query execution time normal, or are queries taking unusually long to respond?

• Is database connection availability stable, or are there signs of connection bottlenecks affecting workload performance?

This section will deal with the first layer of the layer model only, as the Network Layer have been discussed elaborately in the Monitoring Unix and Windows Servers document.