Monitoring GridGain Cluster Health

Losing server nodes reduces cluster capacity and can eventually affect availability. Track whether all expected cluster nodes are present.

Add a Nodes widget to see the status of each node at a glance.

Then add two Metrics (single value) widgets with the following metrics:

These are cluster-wide metrics, so when a widget asks which node to read from, select any node; every node reports the same value. Rename the widgets to Server Nodes and Baseline Nodes respectively. You set up an alert on the cluster’s node count in the alerting part.

Besides monitoring the data stored in the cluster, you should also monitor the JVM running on each node: how much CPU it is using, how much of that CPU goes to garbage collection, and how much Java heap it is using. A rising GC load can be an early sign of heap pressure, well before it surfaces as an out-of-memory error.

Add a Metrics (chart) widget with two metrics:

Rename the widget to CPU & GC Load.

Then add a Metrics (chart) widget on sys.memory.heap.used and rename it to JVM Heap. This is the Java heap GridGain uses, primarily for temporary objects created while processing operations.

GridGain distributes each cache’s partitions across the cluster’s baseline nodes. A Heat Map widget shows how a chosen metric is distributed across those nodes, making it easy to spot an unbalanced cluster.

Add a Heat Map widget on io.dataregion.default.OffheapUsedSize and rename it to Data Distribution. The widget shows one tile per server node, colored on a relative Min–Max scale: nodes holding more data appear "hotter" (toward red), while those holding less data appear "cooler" (toward green). When data is evenly distributed, as in this demo, all tiles have the same color. A noticeably hotter tile indicates that one node holds a disproportionate share of the data.

The color scale compares nodes only with each other. It does not show how full the off-heap memory is. Equal colors indicate balanced data distribution, not a particular level of off-heap memory usage.

With all widgets in place, the Cluster Health dashboard provides a quick overview of your cluster’s health:

This dashboard shows whether the cluster’s infrastructure is healthy. In the next part, you set up alerting, so GridGain notifies you when something that could affect the cluster’s stability or performance happens, even when no one is watching a dashboard: