This topic describes the metrics of FE and BE processes.

View metrics

  • FE metrics: You can view the data of FE metrics by using the URL http://fe_host:fe_http_port/metrics. You can obtain the data of FE metrics in the JSON format by visiting http://fe_host:fe_http_port/metrics?type=json.
  • BE: You can view the data of BE metrics by using the URL http://be_host:be_webserver_port/metrics. You can obtain the data of BE metrics in the JSON format by visiting http://be_host:be_webserver_port/metrics?type=json.
Note
  • By default, FE and BE metrics are displayed in the Prometheus format. For more information, see Prometheus.
  • In the preceding URLs, replace host with the IP address of the master node in your E-MapReduce (EMR) cluster. For BE processes, you must use the private IP address. For FE processes, you can use the private IP address or the public IP address based on your network environment.
  • The default value of http_port is 8031, and the default value of webserver_port is 8041. You can obtain the port numbers by running the show frontends command.

FE metrics

doris_fe_snmp{name="tcp_in_errs"}

This metric value is equivalent to the value of the Tcp: InErrs field in /proc/net/snmp. This metric indicates the number of error Transmission Control Protocol (TCP) packets that are received. The incidence rate can be calculated in combination with the sampling period. This metric is used to troubleshoot network issues.

doris_fe_snmp{name="tcp_retrans_segs"}

This metric value is equivalent to the value of the Tcp: RetransSegs field in /proc/net/snmp. This metric indicates the number of TCP packets that are retransmitted. The incidence rate can be calculated in combination with the sampling period. This metric is used to troubleshoot network issues.

doris_fe_snmp{name="tcp_in_segs"}

This metric value is equivalent to the value of the Tcp: InSegs field in /proc/net/snmp. This metric indicates the number of all received TCP packets. The error rate of received TCP packets can be calculated by using the following formula: (NEW_tcp_in_errs - OLD_tcp_in_errs)/(NEW_tcp_in_segs - OLD_tcp_in_segs). This metric is used to troubleshoot network issues.

doris_fe_snmp{name="tcp_out_segs"}

This metric value is equivalent to the value of the Tcp: OutSegs field in /proc/net/snmp. This metric indicates the number of sent TCP packets that have the RST flag. The retransmission rate of TCP packets can be calculated by using the following formula: (NEW_tcp_tcp_retrans_segs - OLD_tcp_retrans_segs)/(NEW_tcp_out_segs - OLD_tcp_out_segs). This metric is used to troubleshoot network issues.

doris_fe_meminfo{name="memory_total"}

This metric value is equivalent to the value of the MemTotal field in /proc/meminfo. This metric indicates the size of all available memory space, which is equal to the total physical memory size minus the reserved space size and core memory size. This metric is used to troubleshoot memory issues.

doris_fe_meminfo{name="memory_free"}

This metric value is equivalent to the value of the MemFree field in /proc/meminfo. This metric indicates the unused memory size of the system. This metric is used to troubleshoot memory issues.

doris_fe_meminfo{name="memory_available"}

This metric value is equivalent to the value of the MemAvailable field in /proc/meminfo. This metric indicates the available memory size of the system, which is equal to the reclaimable memory size plus the value of the MemFree field. Some occupied memory in the system can be reclaimed. This metric is used to troubleshoot memory issues.

doris_fe_meminfo{name="buffers"}

This metric value is equivalent to the value of the Buffers field in /proc/meminfo. This metric indicates the memory size that is used by block devices to cache metadata and pages of the file system. This metric is used to troubleshoot memory issues.

doris_fe_meminfo{name="cached"}

This metric value is equivalent to the value of the Cached field in /proc/meminfo. This metric indicates the memory size that is allocated to the file buffer. This metric is used to troubleshoot memory issues.

jvm_thread{type="count"}

This metric indicates the total number of Java virtual machine (JVM) threads on an FE node, including Daemon threads and non-Daemon threads. This metric is used to troubleshoot JVM thread operation issues on FE nodes.

jvm_thread{type="peak_count"}

This metric indicates the peak number of threads on an FE node after the JVM started. This metric is used to troubleshoot JVM thread operation issues on FE nodes.

jvm_thread{type="new_count"}

This metric indicates the number of JVM threads in the NEW state on an FE node. This metric is used to troubleshoot JVM thread operation issues on FE nodes.

jvm_thread{type="runnable_count"}

This metric indicates the number of JVM threads in the RUNNABLE state on an FE node. This metric is used to troubleshoot JVM thread operation issues on FE nodes.

jvm_thread{type="blocked_count"}

This metric indicates the number of JVM threads in the BLOCKED state on an FE node. This metric is used to troubleshoot JVM thread operation issues on FE nodes.

jvm_thread{type="waiting_count"}

This metric indicates the number of JVM threads in the WAITING state on an FE node. This metric is used to troubleshoot JVM thread operation issues on FE nodes.

jvm_thread{type="timed_waiting_count"}

This metric indicates the number of JVM threads in the TIMED_WAITING state on an FE node. This metric is used to troubleshoot JVM thread operation issues on FE nodes.

jvm_thread{type="terminated_count"}

This metric indicates the number of JVM threads in the TERMINATED state on an FE node. This metric is used to troubleshoot JVM thread operation issues on FE nodes.

BE metrics

doris_be_snmp{name="tcp_in_errs"}

This metric value is equivalent to the value of the Tcp: InErrs field in /proc/net/snmp. This metric indicates the number of error TCP packets that are received. The incidence rate can be calculated in combination with the sampling period. This metric is used to troubleshoot network issues.

doris_be_snmp{name="tcp_retrans_segs"}

This metric value is equivalent to the value of the Tcp: RetransSegs field in /proc/net/snmp. This metric indicates the number of TCP packets that are retransmitted. The incidence rate can be calculated in combination with the sampling period. This metric is used to troubleshoot network issues.

doris_be_snmp{name="tcp_in_segs"}

This metric value is equivalent to the value of the Tcp: InSegs field in /proc/net/snmp. This metric indicates the number of all received TCP packets. The error rate of received TCP packets can be calculated by using the following formula: (NEW_tcp_in_errs - OLD_tcp_in_errs)/(NEW_tcp_in_segs - OLD_tcp_in_segs) This metric is used to troubleshoot network issues.

doris_be_snmp{name="tcp_out_segs"}

This metric value is equivalent to the value of the Tcp: OutSegs field in /proc/net/snmp. This metric indicates the number of sent TCP packets that have the RST flag. The retransmission rate of TCP packets can be calculated by using the following formula: (NEW_tcp_tcp_retrans_segs - OLD_tcp_retrans_segs)/(NEW_tcp_out_segs - OLD_tcp_out_segs) This metric is used to troubleshoot network issues.

doris_be_compaction_mem_current_consumption

This metric indicates the total MemPool consumption of all Compaction threads. You can use this metric value to quickly identify whether Compaction threads use too much memory. If excessive memory resources are used, the memory usage may be excessively high, or out of memory (OOM) errors may occur. This metric is used to troubleshoot memory usage issues.