[ 
https://issues.apache.org/jira/browse/HDDS-16533?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Sadanand Shenoy reassigned HDDS-16533:
--------------------------------------

    Assignee: Sadanand Shenoy

> Add optional checkpoint download mode to skip compaction artifacts (sst 
> backup + compaction logs) for bootstrap/recovery
> ------------------------------------------------------------------------------------------------------------------------
>
>                 Key: HDDS-16533
>                 URL: https://issues.apache.org/jira/browse/HDDS-16533
>             Project: Apache Ozone
>          Issue Type: Bug
>            Reporter: Sadanand Shenoy
>            Assignee: Sadanand Shenoy
>            Priority: Major
>
> In some OM bootstrap/recovery scenarios, checkpoint transfer is large and 
> expensive because it includes compaction artifacts 
> ({{{}compaction-sst-backup{}}} and compaction logs) in addition to the active 
> OM DB checkpoint and snapshot data.
> This JIRA proposes an opt-in checkpoint download mode to skip transferring 
> compaction artifacts while still transferring the core checkpoint needed for 
> OM bootstrap/install.
> h3. Problem
>  * Checkpoint payload can be very large due to compaction artifact 
> directories.
>  * Large transfer increases bootstrap/recovery time and operational risk 
> during incidents.
>  * Operators need a lighter recovery path when incremental snapshot-diff 
> history is not critical.
> h3. Proposed change
>  * Add an optional flag in checkpoint download/serve flow (CLI + server 
> handling) to skip:
>  ** compaction SST backup directory
>  ** compaction log directory
>  * Keep default behavior unchanged (backward compatible).
>  * Add explicit marker/metadata in checkpoint response/content indicating 
> compaction artifacts were intentionally skipped.
> h3. Expected behavior impact
>  * OM bootstrap/install should still succeed with the reduced checkpoint.
>  * Tradeoff: snapshot diff may become slower after restore because 
> incremental diff history is incomplete; system may fall back to more full 
> diff computations until new history is rebuilt.
> h3.  



--
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