Monitor Metrics

Doris FE process and BE processes provide complete monitoring metrics. Monitoring metrics can be divided into two categories：

1. Process monitoring: mainly displays some monitoring values of the Doris process itself .
2. Node monitoring: mainly displays the monitoring of the node machine itself where the Doris process is located, such as CPU , memory, IO , network , etc.

You can obtain the current monitoring by accessing the http port of the FE or BE node . like :

curl http://fe_host:http_port/metrics
curl http://be_host:webserver_port/metrics

Monitoring metrics are generated in a Prometheus compatible format, for eg:

doris_fe_cache_added{type="partition"} 0
doris_fe_cache_added{type="sql"} 0
doris_fe_cache_hit{type="partition"} 0
doris_fe_cache_hit{type="sql"} 0
doris_fe_connection_total 2

Monitoring metrics in Json format can be fetched using type parameter in rest interface, for eg：

curl http://fe_host:http_port/metrics?type=json
curl http://be_host:webserver_port/metrics?type=json

Monitoring levels and best practices

The last column in the table marks the importance level of the monitoring items. P0 means the most important, the larger the value , the lower the importance.

The vast majority of monitoring metric types are Counter . That is the cumulative value. You can obtain effective information by collecting monitoring values at intervals (such as every 15 seconds) and calculating the slope per unit time .

The query error rate can be obtained by calculating the slope of doris_fe_query_err, presented as (error per second).

FE Monitoring Metrics

Process monitoring

JVM metrics

Machine metrics

BE metrics metrics

Process metrics

Machine metrics