Chi-Hsuan Huang created HDDS-16384:
--------------------------------------

             Summary: Document fileSize bin semantics for the Recon fileCount 
API
                 Key: HDDS-16384
                 URL: https://issues.apache.org/jira/browse/HDDS-16384
             Project: Apache Ozone
          Issue Type: Task
          Components: Ozone Recon
            Reporter: Chi-Hsuan Huang


The Recon GET /api/v1/utilization/fileCount endpoint exposes an optional 
fileSize filter, but the current documentation does not fully describe its 
semantics.

fileSize is matched against the exact upper bound of a file\-size histogram 
bin. It is not an arbitrary raw file size and does not mean a cumulative 
less\-than\-or\-equal threshold. A positive value that is not a bin upper bound 
returns an empty result. The filter can be used independently or together with 
volume and/or bucket.

This became more visible while reviewing HDDS\-16375 and apache/ozone PR 
#11203, which fixes the scan path so that it no longer ignores fileSize when 
volume or bucket is omitted. The new tests also explicitly establish the 
behavior for a value that is not a bin upper bound.

Update the documentation sources in the apache/ozone repository:* 
hadoop\-hdds/docs/content/interface/ReconApi.md
* hadoop\-hdds/docs/content/interface/ReconApi.zh.md
* hadoop\-hdds/docs/themes/ozonedoc/static/swagger\-resources/recon\-api.yaml
* hadoop\-ozone/recon/src/main/resources/chatbot/recon\-tool\-semantics.md

Also clarify the api\_v1\_utilization\_fileCount tool description in 
LlmToolSpecFactory so the Recon chatbot does not interpret fileSize as an 
arbitrary size or cumulative threshold. In particular, review the current 
example wording “how many files are under 1MB”, since the endpoint returns a 
single matching histogram bin rather than a cumulative count.

Related:
\* 
[https://issues.apache.org/jira/browse/HDDS\-16375|https://issues.apache.org/jira/browse/HDDS-16375]
\* 
[https://github.com/apache/ozone/pull/11203|https://github.com/apache/ozone/pull/11203]





--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to