Ilmar Kerm

Oracle, databases, Linux and maybe more

Exporting Oracle Rest Data Services metrics for Prometheus

Written by:
Category: Blog entry
Published:

Tested with ORDS 24.3 running using GraalVM 21.

ORDS documentation has a chapter how to push ORDS metrics to OpenTelemetry endpoint, which also is supported by Prometheus. But Prometheus traditionally is using the opposite method, that Prometheus itself will regularly connect to monitored service endpoints and scrape all its metrics. Similar JavaAgent method can also be deployed to expose ORDS metrics as a traditional Prometheus pull based endpoint.

For this we need to deploy Prometheus JMX exporter as a Java agent in ORDS, this will expose /metrics endpoint that Prometheus can scrape.

First, JMX exporter downloads and documentation is available here.

# Download JMX exporter to ORDS host
curl -o jmx_prometheus_javaagent.jar "https://repo1.maven.org/maven2/io/prometheus/jmx/jmx_prometheus_javaagent/1.0.1/jmx_prometheus_javaagent-1.0.1.jar"

Create configuration file, config.yaml with the following contents. It will format the JMX data into more usable Prometheus metric names.

rules:
    # Reformatting Oracle UCP metrics to have more usable names and adding data types
    - pattern: "oracle.ucp.admin.UniversalConnectionPoolMBean<name=.+, poolName=\\|(.+)\\|(.+)\\|.+><>connectionsClosedCount"
      name: oracle_ucp_connectionsClosedCount
      type: COUNTER
      labels:
          poolName: $1_$2
    - pattern: "oracle.ucp.admin.UniversalConnectionPoolMBean<name=.+, poolName=\\|(.+)\\|(.+)\\|.+><>connectionsCreatedCount"
      name: oracle_ucp_connectionsCreatedCount
      type: COUNTER
      labels:
          poolName: $1_$2
    - pattern: "oracle.ucp.admin.UniversalConnectionPoolMBean<name=.+, poolName=\\|(.+)\\|(.+)\\|.+><>cumulative(\\w+)"
      name: oracle_ucp_cumulative$3
      type: COUNTER
      labels:
          poolName: $1_$2
    - pattern: "oracle.ucp.admin.UniversalConnectionPoolMBean<name=.+, poolName=\\|(.+)\\|(.+)\\|.+><>(\\w+)"
      name: oracle_ucp_$3
      type: GAUGE
      labels:
          poolName: $1_$2
    # This pattern below will add all the rest, tons of detailed java internal things
    # Comment out if you do not want to see them
    - pattern: '.*'

I will assume, that jmx_prometheus_javaagent.jar and config.yaml are placed under /home/ords

Next, change ORDS startup script so it would include the JMX agent. The easiest way is to use environment variable _JAVA_OPTIONS for it.

# Set Startup Java options
# 10.10.10.10 is my local server IP where metrics exporter will bind to, default is localhost
# 21022 is the port JMX exporter will listen to
# With this ORDS metrics would be exposed as http://10.10.10.10:21022/metrics

export _JAVA_OPTIONS="-javaagent:/home/ords/jmx_prometheus_javaagent.jar=10.10.10.10:21022:/home/ords/config.yaml"

# Start ORDS in standalone mode as usual
ords serve

Below is my full ORDS SystemD service file – /etc/systemd/system/ords.service

[Unit]
Description=Oracle Rest Data Services
After=syslog.target network.target

[Service]
Type=simple
User=ords
Group=ords
Restart=always
RestartSec=30
Environment="_JAVA_OPTIONS=-Xms3G -Xmx3G -javaagent:/home/ords/jmx_prometheus_javaagent.jar=10.10.10.10:21022:/home/ords/config.yaml"
Environment="JAVA_HOME=/home/ords/graalvm"
#Environment="JAVA_TOOL_OPTIONS=-Djava.util.logging.config.file=/home/ords/logging.conf"
ExecStart=/home/ords/ords/bin/ords --config /etc/ords/config serve --secure --port 8443 --key /etc/ords/server.key --certificate /etc/ords/server.pem

[Install]
WantedBy=multi-user.target

After restarting ORDS I can query its metrics endpoint.

curl http://10.10.10.10:21022/metrics

# You will see many Java and JVM metrics in the output. Example...
jvm_memory_pool_max_bytes{pool="Compressed Class Space"} 1.073741824E9
jvm_memory_pool_max_bytes{pool="G1 Eden Space"} -1.0
jvm_memory_pool_max_bytes{pool="G1 Old Gen"} 3.221225472E9
jvm_memory_pool_max_bytes{pool="G1 Survivor Space"} -1.0
jvm_memory_pool_max_bytes{pool="Metaspace"} -1.0
jvm_memory_pool_used_bytes{pool="CodeHeap 'non-nmethods'"} 1822336.0
jvm_memory_pool_used_bytes{pool="CodeHeap 'non-profiled nmethods'"} 5918080.0
jvm_memory_pool_used_bytes{pool="CodeHeap 'profiled nmethods'"} 2.3397888E7
jvm_memory_pool_used_bytes{pool="Compressed Class Space"} 7328848.0
jvm_memory_pool_used_bytes{pool="G1 Eden Space"} 2.57949696E8
jvm_memory_pool_used_bytes{pool="G1 Old Gen"} 2.280663304E9
jvm_memory_pool_used_bytes{pool="G1 Survivor Space"} 8528.0
jvm_memory_pool_used_bytes{pool="Metaspace"} 6.750048E7
jvm_memory_used_bytes{area="heap"} 2.538621528E9
jvm_memory_used_bytes{area="nonheap"} 1.05967632E8
jvm_threads_deadlocked_monitor 0.0
jvm_threads_peak 62.0
jvm_threads_started_total 62.0
jvm_threads_state{state="BLOCKED"} 0.0
jvm_threads_state{state="NEW"} 0.0
jvm_threads_state{state="RUNNABLE"} 12.0
jvm_threads_state{state="TERMINATED"} 0.0
jvm_threads_state{state="TIMED_WAITING"} 20.0
jvm_threads_state{state="UNKNOWN"} 0.0
jvm_threads_state{state="WAITING"} 15.0

# ORDS database connection pool metrics will be exported like this
# Just an example... all UCP attributes are exported, for all ORDS connection pools

oracle_ucp_abandonedConnectionTimeout{poolName="backoffice_lo"} 0.0
oracle_ucp_abandonedConnectionTimeout{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_abandonedConnectionsCount{poolName="backoffice_lo"} 0.0
oracle_ucp_abandonedConnectionsCount{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_availableConnectionsCount{poolName="backoffice_lo"} 10.0
oracle_ucp_availableConnectionsCount{poolName="marketing_communications_2_lo"} 10.0
oracle_ucp_averageBorrowedConnectionsCount{poolName="backoffice_lo"} 1.0
oracle_ucp_averageBorrowedConnectionsCount{poolName="marketing_communications_2_lo"} 1.0
oracle_ucp_averageConnectionWaitTime{poolName="backoffice_lo"} 0.0
oracle_ucp_averageConnectionWaitTime{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_borrowedConnectionsCount{poolName="backoffice_lo"} 0.0
oracle_ucp_borrowedConnectionsCount{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_bufferSize{poolName="backoffice_lo"} 1024.0
oracle_ucp_bufferSize{poolName="marketing_communications_2_lo"} 1024.0
oracle_ucp_connectionHarvestMaxCount{poolName="backoffice_lo"} 1.0
oracle_ucp_connectionHarvestMaxCount{poolName="marketing_communications_2_lo"} 1.0
oracle_ucp_connectionHarvestTriggerCount{poolName="backoffice_lo"} 2.147483647E9
oracle_ucp_connectionHarvestTriggerCount{poolName="marketing_communications_2_lo"} 2.147483647E9
oracle_ucp_connectionRepurposeCount{poolName="backoffice_lo"} 0.0
oracle_ucp_connectionRepurposeCount{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_connectionValidationTimeout{poolName="backoffice_lo"} 15.0
oracle_ucp_connectionValidationTimeout{poolName="marketing_communications_2_lo"} 15.0
oracle_ucp_connectionWaitTimeout{poolName="backoffice_lo"} 3.0
oracle_ucp_connectionWaitTimeout{poolName="marketing_communications_2_lo"} 3.0
oracle_ucp_connectionsClosedCount{poolName="backoffice_lo"} 0.0
oracle_ucp_connectionsClosedCount{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_connectionsCreatedCount{poolName="backoffice_lo"} 10.0
oracle_ucp_connectionsCreatedCount{poolName="marketing_communications_2_lo"} 10.0
oracle_ucp_createConnectionInBorrowThread{poolName="backoffice_lo"} 1.0
oracle_ucp_createConnectionInBorrowThread{poolName="marketing_communications_2_lo"} 1.0
oracle_ucp_cumulativeConnectionBorrowedCount{poolName="backoffice_lo"} 1.0
oracle_ucp_cumulativeConnectionBorrowedCount{poolName="marketing_communications_2_lo"} 1.0
oracle_ucp_cumulativeConnectionReturnedCount{poolName="backoffice_lo"} 1.0
oracle_ucp_cumulativeConnectionReturnedCount{poolName="marketing_communications_2_lo"} 1.0
oracle_ucp_cumulativeConnectionUseTime{poolName="backoffice_lo"} 60.0
oracle_ucp_cumulativeConnectionUseTime{poolName="marketing_communications_2_lo"} 30.0
oracle_ucp_cumulativeConnectionWaitTime{poolName="backoffice_lo"} 0.0
oracle_ucp_cumulativeConnectionWaitTime{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_cumulativeFailedConnectionWaitCount{poolName="backoffice_lo"} 0.0
oracle_ucp_cumulativeFailedConnectionWaitCount{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_cumulativeFailedConnectionWaitTime{poolName="backoffice_lo"} 0.0
oracle_ucp_cumulativeFailedConnectionWaitTime{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_cumulativeSuccessfulConnectionWaitCount{poolName="backoffice_lo"} 0.0
oracle_ucp_cumulativeSuccessfulConnectionWaitCount{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_cumulativeSuccessfulConnectionWaitTime{poolName="backoffice_lo"} 0.0
oracle_ucp_cumulativeSuccessfulConnectionWaitTime{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_failedAffinityBasedBorrowCount{poolName="backoffice_lo"} 0.0
oracle_ucp_failedAffinityBasedBorrowCount{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_failedRCLBBasedBorrowCount{poolName="backoffice_lo"} 0.0
oracle_ucp_failedRCLBBasedBorrowCount{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_failoverEnabled{poolName="backoffice_lo"} 0.0
oracle_ucp_failoverEnabled{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_inactiveConnectionTimeout{poolName="backoffice_lo"} 1800.0
oracle_ucp_inactiveConnectionTimeout{poolName="marketing_communications_2_lo"} 1800.0
oracle_ucp_initialPoolSize{poolName="backoffice_lo"} 10.0
oracle_ucp_initialPoolSize{poolName="marketing_communications_2_lo"} 10.0
oracle_ucp_labeledConnectionsCount{poolName="backoffice_lo"} 0.0
oracle_ucp_labeledConnectionsCount{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_loggingEnabled{poolName="backoffice_lo"} 0.0
oracle_ucp_loggingEnabled{poolName="marketing_communications_2_lo"} 0.0
oracle_ucp_maxConnectionReuseCount{poolName="backoffice_lo"} 1000.0

Tags:

Ilmar Kerm