Multi-Instance Monitoring
On the multi-instance monitoring screen, you can gather the metrics of multiple nodes in a cluster into one dashboard for comparison. In a distributed environment, load is sometimes concentrated on a specific node, and placing per-node metrics side by side lets you quickly find the skew.
Basic screen guide
The default dashboard (Default preset) consists of charts that gather the key metrics of each node and an active session table.
Narrowing the target by cluster
Select Agent by cluster in the agent selector at the top of the screen to show the cluster list. Select one cluster to leave only the nodes of that cluster on the dashboard. This menu appears only for products that have the concept of a cluster.
Default dashboard layout
The default board places three resource widgets (CPU, Memory, Disk Usage) in one row, with the active session table below them. Network Bandwidth has been removed from the default board. It remains in the widget palette, so you can add it if needed.
The bars in the resource widgets are sorted by cluster, role, and then name. Sorting by role alone would mix nodes from different clusters, making it hard to tell which group a node belongs to.
Host / Node switch in the Memory widget
Memory can be measured by OS accounting or by engine accounting. The denominator is the same physical memory (host_total_mb); only the numerator changes.
| Item | Description |
|---|---|
Host | Total usage of the machine (host_used_mb) |
Node | This node's memsqld usage within it (memsql_used_mb) |
Choose with the Host and Node buttons in the widget header. The widget title also changes to Host Memory or Node Memory with your choice, so the title alone tells you which accounting you are looking at. The selection is kept only within your screen and is not saved to the dashboard, so it does not change other users' screens on a shared board.
The two values can actually differ. Memory used by a process includes the page cache, which is excluded from the host usage calculation, so a node may show a larger Node value than Host.
Empty when a value is missing
If either the numerator or the denominator is missing, the gauge is left uncolored and the caption shows a dash (-). Previously it silently fell back to a ratio on a different basis, but that was removed because a value with a different denominator appearing in the same place is misleading.
Widget description icon
Hover over a resource widget to see the description icon. Hover over the icon to see, on the first line, which system view the value was read from. Metrics with the same name can have different values depending on the view.
| Widget | View | Metric |
|---|---|---|
CPU | MV_SYSINFO_CPU | CPU usage |
Memory | MV_SYSINFO_MEM | Host used / Physical, memsqld / Physical |
Disk Usage | MV_SYSINFO_DISK | Mount used / Capacity |
Network Bandwidth | MV_SYSINFO_NET | IN (received) · OUT (sent) |
Captions and guidance text show human-readable names instead of collection field names. For example, host_used_mb is shown as Host used, and os_mount_total_mb as Capacity.
The Idle item is no longer shown in the CPU widget. The gauge already represents the usage rate, so it amounted to writing the value subtracted from 100 one more time.
Adding widgets
You can compose the multi-instance dashboard by adding SingleStore-specific widgets from the widget palette. The method for adding, editing, and deleting widgets is the same as the common database dashboard editing method.
Active session area
In the active session table at the bottom of the dashboard, you can check sessions across the cluster. For column descriptions and the SQL detail view, see the Instance Monitoring and Top SQL documents.