Skip to main content
Version: Next

Recon Web UI

Introduction​

Apache Ozone Recon is a read-only monitoring and observability service for Apache Ozone clusters, designed for cluster administrators. Recon aggregates metadata from Ozone Manager (OM), Storage Container Manager (SCM), and Datanodes, and exposes this information through a web-based user interface.

Recon does not modify cluster state. It is designed to provide visibility, diagnostics, and insight into the health and usage of an Ozone cluster.

Accessing the Recon UI​

The Recon UI is available via the Recon HTTP endpoint, typically:

http://<recon-host>:9888/

The left navigation panel provides access to all available views:

  • Overview
  • Volumes
  • Buckets
  • Datanodes
  • Pipelines
  • Containers
  • Insights
  • OM DB Insights
  • Namespace Usage

The Cluster Capacity page is reached from the View More link on the Overview page rather than from the left navigation.

Common conventions​

Most Recon UI pages share the same controls and behaviors. They are described once here; each page notes only where it differs.

  • Auto Reload panel — an Auto Refresh toggle that re-queries Recon every 60 seconds (and remembers your choice for the session) plus a manual reload button showing the last refreshed time. Auto Refresh only re-queries Recon; it does not force Recon to re-sync from OM. Some pages (for example Insights, Heatmap, and OM DB Insights) have no Auto Reload panel.
  • Data freshness — namespace data (volumes, buckets, keys, and their summaries) comes from Recon's own copy of the Ozone Manager (OM) database, updated by a periodic sync; datanode, container, pipeline, and capacity figures come from datanode reports and heartbeats. Values are only as current as the last sync or report. The Overview page shows the last DB Synced at time and can trigger an on-demand OM sync.
  • Columns selector — table pages let you choose which columns are visible; a key identifier column is always pinned.
  • Limit selector — on large listings, controls how many rows are fetched from the server (options such as 1000/5000/10000/20000; default 1000). Search, sorting, and filters apply to the loaded rows only, not to rows beyond the limit — raise the limit if you expect more.
  • Error states — on a fetch error, a page reports a data-fetch error and its table stays empty; until the first successful load, counts may appear as zero or N/A.

AI Assistant​

Recon includes an optional AI Assistant that lets you query cluster metadata in plain English from within the UI. It is disabled by default and requires operator configuration. See the Recon AI Assistant page for setup, configuration, security, and usage guidance.