[ 
https://issues.apache.org/jira/browse/HDDS-15614?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18090049#comment-18090049
 ] 

Ivan Andika edited comment on HDDS-15614 at 6/19/26 2:23 AM:
-------------------------------------------------------------

FYI, we still have our internal use case for pull-based replication. Internally 
we are developing a pull based asynchronous write pipeline between the main 
write pipeline and a remote DC datanode, the remote DC datanode can pull from 
any of these 3 datanodes in the main pipeline (first incrementally in PutBlock 
boundary and using a fallback copy if incremental replication fails). Using 
pull-based is easier since the remote DC datanode can simply failover to other 
datanodes should one fail. If we use push, then we need further coordination 
(e.g SCM to detect the source DN failure and send another push command to 
another source DN) if the source DN fails.


was (Author: JIRAUSER298977):
FYI, we still have our use case for pull-based replication. Internally we are 
developing a pull based asynchronous write pipeline between the main write 
pipeline and a remote DC datanode, the remote DC datanode can pull from any of 
these 3 datanodes in the main pipeline. Using pull-based is easier since the 
remote DC datanode can simply failover to other datanodes should one fail. If 
we use push, then we need further coordination if the source DC fails.

> Remove Datanode download/pull container replication
> ---------------------------------------------------
>
>                 Key: HDDS-15614
>                 URL: https://issues.apache.org/jira/browse/HDDS-15614
>             Project: Apache Ozone
>          Issue Type: Improvement
>          Components: Ozone Datanode
>            Reporter: Ethan Rose
>            Assignee: Ethan Rose
>            Priority: Major
>
> Datanodes have used push replication to move containers by default since 
> early 2023. There is no reason to continue using the pull model which is now 
> disabled behind a feature flag. We can safely remove it, which will be easier 
> to do before ZDU is merged into master.



--
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