bharatviswa504 commented on pull request #2151:
URL: https://github.com/apache/ozone/pull/2151#issuecomment-822195401


   > > @rakeshadr
   > > I think with this approach, a cluster upgraded from an older version and 
if someone want to try a new layout version format, they cannot use it. IMHO, I 
think we should support older buckets read/write, and provide the capability to 
support users to try out the new format on the newly created buckets.
   > > One Idea here is to persist format layout in bucket metadata and use 
that. And also iterate the buckets(Even now we iterate buckets during startup) 
and set layout version of the buckets to older format layout during startup if 
the format is not set. (I think this approach has been discussed during initial 
meetings)
   > 
   > Thanks @bharatviswa504 for bringing the point. I think, I haven't explain 
it clearly about the purpose of this patch. I will try to add more details here:
   > 
   > Yes, I have persisted the FSO layout in `bucketInfo.getMetadata()` (bucket 
level property) to differentiate older buckets and newer buckets. As OM is 
persisting the layout in the bucket level the feature can be extended to 
support both older and newer buckets. Please see rename and delete APIs, the 
older and newer buckets are implemented in that fashion. I have plans to 
support all the cases gracefully, I hope [my comment in the 
jira](https://issues.apache.org/jira/browse/HDDS-5094?focusedCommentId=17319502&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17319502)
 conveys the same thoughts.
   > 
   > The purpose of this patch is to validate and fail during OM startup till I 
support and implement all the cases incrementally(phase by phase), then will 
remove this check. This means that, now in phase-1 the new feature is allowed 
only in a fresh new cluster. Later, I agree to support this case in follow-up 
tasks, existing old key(s) should be visible in prefix based OM layout.
   
   Thank You @rakeshadr for the clear explanation and offline discussion. I am 
fine with this, if we are planning to address them in furthere Jiras.


-- 
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.

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