Uploaded image for project: 'WildFly'
  1. WildFly
  2. WFLY-20799

Loading Prometheus Metrics takes up to 5 Minutes due to org.jboss.as.jpa.hibernate.management.QueryName

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Unresolved
    • Icon: Major Major
    • None
    • 36.0.1.Final
    • JPA / Hibernate, Metrics
    • None
    • ---
    • ---

      After upgrading from Wildfly 25.0.1.Final to 36.0.1.Final we noticed 1 CPU core being always active.

      The attached flamegraph from an idle environment shows that all CPU time is spend in PrometheusExporter.export. More precise the gathering of the JPA metrics consumes all the CPU time.

      We pull the metrics from the /metrics endpoint on port 9990 every minute.

      The time to request the metrics is about 10 seconds after a server start and increases after some load to a max time of 5 Minutes (as far as I have seen).

      We use the default values for the metrics as they come in the standalone.xml.

        1. idle.zip
          8 kB
        2. metrics-query-name-improvement.zip
          10 kB

          1.
          		 Improve the performance of the org.jboss.as.jpa.hibernate.management.QueryName class Sub-task Resolved Major Scott Marlow
          2.
          		 Default to a lower hibernate.statistics.query_max_size value of 200 unless application overrides it Sub-task Pull Request Sent Major Scott Marlow
          3.
          		 optimize the JPA/Persistence Statistics SPI to better handle higher hibernate.statistics.query_max_size settings Sub-task Open Major Scott Marlow

              smarlow1@redhat.com Scott Marlow
              rschimpf Robin Schimpf
              Votes:
              0 Vote for this issue
              Watchers:
              4 Start watching this issue

                Created:
                Updated: