Monitoring Apache IoTDB Database

Monitor the running status of the Apache IoTDB Internet of Things time series database (JVM-related), memory task clusters and other related Metrics.

Operation before monitoring

You need to enable the metrics function in IoTDB, which will provide interface data in the form of prometheus metrics.

To enable the metrics function, refer to Official Documentation

The main steps are as follows:

1. The metric collection is disabled by default, you need to modify the parameters in conf/iotdb-metric.yml first, then restart the server

   # Whether to start the monitoring module, the default is false
   enableMetric: true
   
   # Whether to enable operation delay statistics
   enablePerformanceStat: false
   
   # Data provision method, externally provide metrics data through jmx and prometheus protocol, optional parameters: [JMX, PROMETHEUS, IOTDB], IOTDB is closed by default.
   metricReporterList:
      - JMX
      - PROMETHEUS
   
   # The metric architecture used at the bottom layer, optional parameters: [MICROMETER, DROPWIZARD]
   monitorType: MICROMETER
   
   # Initialize the level of the metric, optional parameters: [CORE, IMPORTANT, NORMAL, ALL]
   metricLevel: IMPORTANT
   
   # Predefined metrics set, optional parameters: [JVM, LOGBACK, FILE, PROCESS, SYSTEM]
   predefinedMetrics:
      - JVM
      - FILE

2. Restart IoTDB, open a browser or use curl to access http://servier_ip:9091/metrics, and you can see the metric data.

3. Add the corresponding IoTDB monitoring in HertzBeat.

Configuration parameters

Parameter name	Parameter help description
Monitoring Host	The peer IPV4, IPV6 or domain name to be monitored. Note ⚠️Without protocol header (eg: https://, http://).
Monitoring name	The name that identifies this monitoring, and the name needs to be unique.
Port	The port provided by the IoTDB Metric interface, which is 9091 by default.
Timeout	HTTP request query timeout
Acquisition Interval	Interval time for monitoring periodic data collection, in seconds, the minimum interval that can be set is 30 seconds
Whether to detect	Whether to detect and check the availability of monitoring before adding monitoring, and the operation of adding and modifying will continue after the detection is successful
Description Remarks	More remark information to identify and describe this monitoring, users can remark information here

Collect metrics

Metric collection: cluster_node_status

Metric Name	Metric Unit	Metric Help Description
name	None	Node name IP
status	None	Node status, 1=online 2=offline

Metric collection: jvm_memory_committed_bytes

Metric Name	Metric Unit	Metric Help Description
area	none	heap memory or nonheap memory
id	none	memory block
value	MB	The memory size currently requested by the JVM

Metric collection: jvm_memory_used_bytes

Metric Name	Metric Unit	Metric Help Description
area	none	heap memory or nonheap memory
id	none	memory block
value	MB	JVM used memory size

Metric collection: jvm_threads_states_threads

Metric Name	Metric Unit	Metric Help Description
state	none	thread state
count	None	The number of threads corresponding to the thread state

Index collection: quantity business data

Metric Name	Metric Unit	Metric Help Description
name	None	Business name timeSeries/storageGroup/device/deviceUsingTemplate
type	none	type total/normal/template/template
value	None	The current timeSeries/storageGroup/device/The number of devices that have activated the template

Metric collection: cache_hit cache

Metric Name	Metric Unit	Metric Help Description
name	None	Cache name chunk/timeSeriesMeta/bloomFilter
value	%	chunk/timeSeriesMeta cache hit rate, bloomFilter interception rate

Metric collection: queue task queue

Metric Name	Metric Unit	Metric Help Description
name	None	Queue name compaction_inner/compaction_cross/flush
status	none	status running/waiting
value	None	Number of tasks at current time

Metric collection: thrift_connections

Metric Name	Metric Unit	Metric Help Description
name	None	name
connection	none	thrift current connection number