SingleStore Monitoring
SingleStore is a distributed SQL database that stores data across multiple nodes. As the number of nodes grows, it becomes harder to tell at a glance which node is under load or whether data is skewed to one side. WhaTap database monitoring gathers the overall cluster state, per-node metrics, and SQL performance into one view. It helps DBAs and data platform engineers who operate SingleStore quickly locate load points and data skew.
WhaTap works by having the DBX agent connect to the SingleStore cluster and collect metrics. When you install the agent on only one aggregator node, it automatically detects leaf nodes through MV_NODES and collects the entire cluster, so you do not need to install the agent on every node. For details on how to get started, see the Getting Started document.
SingleStore cluster architecture
A SingleStore cluster is made up of nodes with different roles. Node roles appear throughout the screens, including the ROLE column in the instance list, so it helps to understand them first.
| ROLE | Name | Description |
|---|---|---|
MA | Master Aggregator | The primary aggregator responsible for managing cluster metadata and routing queries |
CA | Child Aggregator | A secondary aggregator that helps distribute query processing |
Leaf | Leaf | A leaf node that stores actual data and performs computation |
In simple terms, the aggregators (MA, CA) are the front desk that receives requests and distributes them, while the leaf nodes are closer to the workers that hold the data and do the work.
Monitoring screen layout
SingleStore monitoring provides screens focused on the items most often checked in a distributed environment.
| Screen | Description |
|---|---|
| Instance Monitoring | Check per-node CPU, memory, disk, and network metrics and active sessions in real time |
| Multi-Instance Monitoring | Gather metrics from multiple nodes into one dashboard for comparison |
| Storage Engine | Check Rowstore and Columnstore memory and disk usage and data skew |
| Partition | Check whether table partitions are evenly distributed across nodes by their deviation |
| Pipeline | Check the throughput, latency, and errors of data ingestion pipelines |
| Database Size | Check capacity trends by database |
| Top SQL | Analyze high-load SQL with metrics and trend charts |
| Session History | Look up past active sessions in chronological order |
| Lock Tree | Check lock waits and deadlocks in a tree view |
Supported environments
Based on the official installation guide, the following environments are supported.
- SingleStore 9.0 or later
- DBX agent installed on an aggregator node (leaf nodes are auto-detected through
MV_NODES) - Monitoring account privileges
grant select, process, cluster on *.*