devmadhuu opened a new pull request, #12062:
URL: https://github.com/apache/ozone/pull/12062

   ## What changes were proposed in this pull request?
   Overall datanode usage can hide storage-type imbalance. For example, two 
datanodes can both be 50% full while one has SSD at 90% and DISK at 10%, and 
the other has the opposite usage.
   
   This PR makes the running ContainerBalancer select moves separately for each 
storage type. SSD replicas move between SSD volumes, and DISK replicas move 
between DISK volumes.
   
   **The changes:**
   
   - Use per-storage-type utilization to identify sources and targets.
   - Take turns scheduling moves across types and rotate which type starts each 
iteration.
   - Share iteration byte limits, per-datanode incoming/outgoing limits and the 
distinct-datanode limit across all types.
   - Cache container candidates separately for each source and storage type.
   - Use the reported volume type when a legacy replica has no recorded type. 
Skip typed moves when both types are unknown or when recorded and volume types 
conflict.
   - Revalidate the source type before dispatch and pass the selected type 
through the existing low-priority copy API.
   - Retain aggregate balancing when no type breakdown is available, and stop 
before selecting typed moves when reports are incomplete.
   
   Existing health, placement, timeout and copy-before-delete checks remain in 
place. Typed balancing requires datanodes that support storage-type copy 
commands.
   
   ## What is the link to the Apache JIRA
   [HDDS-16653 
(https://issues.apache.org/jira/browse/HDDS-16653)](https://issues.apache.org/jira/browse/HDDS-16653)
   
   ## How was this patch tested?
   **Local validation passed:**
   - 579 SCM tests covering balancing, selection, node limits, move handling 
and ReplicationManager.
   - 25 analyzer/estimator tests.
   - 26 CLI tests.
   - 30 datanode copy-command and importer tests.
   - Compilation, test compilation, Checkstyle and PMD.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


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

Reply via email to