Uploaded image for project: 'OpenShift Service Mesh'
  1. OpenShift Service Mesh
  2. OSSM-12036

Pre-compute health values for mesh objects

XMLWordPrintable

    • Icon: Epic Epic
    • Resolution: Done
    • Icon: Undefined Undefined
    • None
    • None
    • Kiali
    • None
    • Pre-compute health values for mesh objects
    • False
    • Hide

      None

      Show
      None
    • False
    • Release Notes
    • In Progress
    • Hide
      **Health Status Pre-Compute and Caching**

      Kiali v2.22 introduces health status pre-compute and caching. Enabled by default. Production mesh sizes are growing and Kiali render times have been increasing, particularly for the Overview and List pages. In response, Kiali v.22 changes its approach to health status calculation. In prior versions Kiali calculated health "on-demand", based on the user's selected duration, configuration settings, and other information, such as pod status. Starting in v.22 Kiali will pre-calculate health status using a single, configurable duration, 5 minutes by default. The cached values increase the responsiveness of the Overview and List pages. Other pages, such as the Traffic graph and Detail pages will continue to calculate health status on-demand, and based on the user's selected duration. Users may notice that the Duration Dropdown selector has been removed from the Overview and List pages.

      Users may notice an increase in backend resource utilization, as the Kiali server will now be calculating and refreshing health status, independent of user sessions. The Kiali CR introduces the following new configuration:

      `spec.health_config.compute.duration: 5m`
      `spec.health_config.compute.refresh_interval: 3m`
      `spec.health_config.compute.TIMEOUT: 10m`

      `spec.kiali_internal.health_cache.enabled: false`

      It is recommended to keep the health cache enabled, as not all features will fall back to on-demand calculation.
      Show
      **Health Status Pre-Compute and Caching** Kiali v2.22 introduces health status pre-compute and caching. Enabled by default. Production mesh sizes are growing and Kiali render times have been increasing, particularly for the Overview and List pages. In response, Kiali v.22 changes its approach to health status calculation. In prior versions Kiali calculated health "on-demand", based on the user's selected duration, configuration settings, and other information, such as pod status. Starting in v.22 Kiali will pre-calculate health status using a single, configurable duration, 5 minutes by default. The cached values increase the responsiveness of the Overview and List pages. Other pages, such as the Traffic graph and Detail pages will continue to calculate health status on-demand, and based on the user's selected duration. Users may notice that the Duration Dropdown selector has been removed from the Overview and List pages. Users may notice an increase in backend resource utilization, as the Kiali server will now be calculating and refreshing health status, independent of user sessions. The Kiali CR introduces the following new configuration: `spec.health_config.compute.duration: 5m` `spec.health_config.compute.refresh_interval: 3m` `spec.health_config.compute.TIMEOUT: 10m` `spec.kiali_internal.health_cache.enabled: false` It is recommended to keep the health cache enabled, as not all features will fall back to on-demand calculation.
    • Feature

      As part of our scalability effort, we want to pre-compute and cache more information about the mesh. Health information is typically generated on-demand, and is often responsible for slow page renders when a lot of health information being accumulated. This very much applies to the Overview and List pages.

      By pre-computing the information we should be able to respond quickly to the client requests, improving the user experience.

      This work can be tracked upstream via https://github.com/kiali/kiali/issues/8900

              rhn-engineering-jshaughn Jay Shaughnessy
              rhn-engineering-jshaughn Jay Shaughnessy
              Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

                Created:
                Updated:
                Resolved: