JMX Exporter returns metrics as follows: To support other metrics, please modify the code or submit issues.
# HELP full_gc_count Exposes JVM's JMX counting the number of Full GCs since the JVM started.
# TYPE full_gc_count counter
full_gc_count 469
# HELP jvm_ThreadCount Total threads in JVM business operations, excluding GC threads.
# TYPE jvm_ThreadCount counter
jvm_ThreadCount 512
Fill in the IP and JMX port 1099 of the JVM nodes to be monitored in targets, e.g.: 172.16.10.3:1099
Prometheus alert rules should be configured according to your own business needs: e.g., the number of Full GCs should not increase more than once in an hour, the number of JVM threads should not exceed a specified value, etc.
In the event of an abnormal alert, you can call remote dump and then execute the appropriate strategy for service repair. Dumping is to capture the scene. Before dumping, you can cut off the abnormal JVM node business traffic, or you can cut after dumping because it may not be possible to capture the scene without business traffic.
Remote dump and execution of JVM node fault strategy are not implemented in this component.
Framework and Metric Description
The code uses the Gin framework to expose metrics.
It does not use the official Prometheus client_golang.
Special Attention
Please read the configuration file content: config.yaml carefully before use, and modify according to the configuration file instructions.