TDinsight is TDengine’s monitoring solution built on Grafana. It gives operations teams a unified web interface for viewing cluster status, node health, request performance metrics, and resource usage in real time. The solution ships with prebuilt dashboards and alerting rules, ready to use after a short setup process.
Prerequisites
Before installing TDinsight, confirm the following components are in place:
- TDengine version 3.0.0.0 or later
- taosAdapter service installed and running
- taosKeeper component installed correctly
- Grafana version 8.0 or later
Installation and configuration
Data Source Plugin
Install the TDengine data source plugin for Grafana through either the command line or the Grafana UI plugin marketplace. Once installed, navigate to the Configuration page in Grafana, add a new Data Source of type TDengine, and fill in the connection details for your TDengine cluster.
Dashboard Import
Import the monitoring dashboards through Grafana’s Dashboard import function. The official JSON configuration file contains all panels, metrics, and layout definitions. After import, the dashboards begin populating with data as soon as the data source connection is active.
Dashboard Features
Cluster Status Overview. Displays the First EP address, version number, active connection count, and status of each node in the cluster. This panel provides a single-page summary of overall cluster health.
DNodes Node Management. Shows detailed information for every data node. Nodes in an abnormal state are highlighted with a distinct color so operators can spot issues immediately.
MNode Management Node Monitoring. Tracks the count and distribution of management nodes across the cluster, helping verify that the control plane is properly distributed.
Request Performance Statistics. Charts the execution counts and latency for Select queries, Delete operations, and Insert writes. This panel helps identify which operation types consume the most resources.
Table-Level Statistics. Tracks count trends for Supertables and regular tables over time, useful for capacity planning and detecting unexpected schema changes.
Resource Usage Monitoring. Shows CPU utilization, memory consumption, disk I/O, and network I/O across cluster nodes.
Alerting Rules
TDinsight includes 14 predefined alerting rules:
| Rule | Trigger |
|---|---|
| High CPU load | CPU usage exceeds threshold |
| Memory usage exceeded | Memory consumption above limit |
| Low disk capacity | Disk space falls below threshold |
| Abnormal disk I/O | Disk I/O deviates from baseline |
| License nearing expiration | License validity period ending soon |
| Node offline | A node becomes unreachable |
| Abnormal request failure rate | Failure rate exceeds threshold |
| Connection count exceeded | Active connections above limit |
| Query response timeout | Query latency exceeds threshold |
| Write performance degradation | Insert throughput drops |
| MNode election anomaly | Management node election issues |
| DNode status change | Data node state transition |
| Slow query | Query exceeds time threshold |
| Heartbeat timeout | Inter-node heartbeat lost |
Each rule supports custom thresholds. Set severity levels (Critical, Warning, Info) based on the business impact of each metric.
Custom alerting rules can be added alongside the defaults. Define them based on your specific operational requirements and the metrics most relevant to your workload patterns.
Troubleshooting
Monitoring data not displayed. Check that both taosAdapter and taosKeeper services are running. Confirm the data source configuration in Grafana points to the correct endpoint. Examine the logs for connection errors or authentication failures.
Alerts not triggering. Verify that the alerting rule configuration is correct and enabled. Check that the notification channel is properly configured. Confirm that metric collection latency is not causing data to arrive after the evaluation window closes.
Summary
Deploying TDinsight early in the production lifecycle shifts operations from reactive to proactive. With real-time visibility into cluster health, resource consumption, and query performance, teams can detect and address issues before they affect users.
