Instance List
The instance list screen gathers the nodes in a SingleStore cluster into a table, one per row. Because you can check the aggregator and leaf nodes together on one screen, it serves as the starting point for quickly scanning the overall cluster state and drilling straight into a problematic node.
Summary area
The area above the table shows the total node count and a breakdown by state.
| Item | Description |
|---|---|
| Total | Total number of nodes being collected |
| Critical | Number of nodes classified as critical |
| Warning | Number of nodes classified as warning |
| Normal | Number of nodes in a normal state |
Node list table
The list is grouped by cluster. Only the Master Aggregator row, which represents the cluster, has an expand button; expanding it shows the nodes belonging to that cluster below. Hover over the expand button to see the cluster name.
Within a cluster, nodes are placed in role order: Master Aggregator, Child Aggregator, and then Leaf. Leaf nodes, which have no hierarchy among themselves, are ordered by server name. In other words, the list order follows the cluster structure rather than the server name.
The default columns are as follows.
| Item | Description |
|---|---|
| Status | The collection and state of the node |
| Server Name | Node name (for example, 1-MA-150, LF-152). Select it to go to the corresponding Instance Monitoring |
| IP | Node IP address |
| Port | Connection port (default 3306) |
| Database | List of databases the node is responsible for |
| ROLE | Node role badge. MA (Master Aggregator), CA (Child Aggregator), Leaf |
CPU (%) | CPU usage trend of the node |
Node Memory (%) | This node's memsqld usage divided by physical memory |
Disk (%) | Mount usage ratio |
Threads_running | Number of running threads |
Resource columns draw only a trend line in the cell, so the unit is indicated by the (%) notation in the header.
The column set has changed
The Questions, innodb_buffer_pool_read_requests, and lock_wait_sessions columns have been removed from the default table. innodb_buffer_pool_read_requests was always empty because the agent does not send a value, and the other two are MySQL protocol compatibility counters that do not represent the state of a distributed engine node.
The Cluster Name column is also hidden in the default table. The table is already grouped by cluster, so the same value repeats on every row. You can turn it back on at any time in Column selection.
For SingleStore, when you install the DBX agent only on an aggregator node, it auto-detects leaf nodes through MV_NODES and the entire cluster appears in the list. You can quickly distinguish roles by the MA, CA, and LF prefixes in node names.
Working with the table
- Column selection: You can add or hide the columns shown in the table.
- Multi-view: Select multiple nodes to switch to a screen that compares them together.
- After selecting a node, you can drill into the detail screen to examine its metrics and sessions in detail.
Data interpretation guide
| Checkpoint | Description |
|---|---|
Sudden spike in Threads_running on a specific node | Suspect load or contention concentrated on that node |
Accumulating lock_wait_sessions | Lock waits are occurring; trace the cause with Lock Tree |
| Metric deviation among Leaf nodes | Possible data skew; check the layout in Partition |