[ 
https://issues.apache.org/jira/browse/KUDU-3466?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17713374#comment-17713374
 ] 

Alexey Serbin commented on KUDU-3466:
-------------------------------------

[~c8679...@163.com],

Kudu has had ability to scan data at a snapshot from the very beginning since 
it uses MVCC.  You could check the following client API:
* C++: 
https://kudu.apache.org/cpp-client-api/classkudu_1_1client_1_1KuduScanner.html#a5b36a405daf09399438d5501b25b9f9f
* Java: 
https://kudu.apache.org/apidocs/org/apache/kudu/client/AbstractKuduScannerBuilder.html#snapshotTimestampMicros-long-

Would be that enough for your use case?

As for difference between two snapshots, that functionality seems to be present 
since backup/restore was implemented, but it's not exposed via client API.  
That's used for incremental backup/restore in Kudu.  Is it something that might 
be useful in your use case?  If so, what sort of client API you'd like to have 
exposed for that purpose?

Thanks!

> I hope kudu can support snapshot management like iceberg, can create, query, 
> expiration snapshots, support query data changes between two snapshots. 
> -----------------------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: KUDU-3466
>                 URL: https://issues.apache.org/jira/browse/KUDU-3466
>             Project: Kudu
>          Issue Type: New Feature
>          Components: api, cfile
>            Reporter: sky
>            Priority: Major
>
> I hope kudu can support snapshot management like iceberg, can create, query, 
> expiration snapshots, support query data changes between two snapshots. This 
> is very necessary, when we write data to kudu in real time, often expect the 
> data to be like the past, query the data before a fixed time, or query the 
> changes between two snapshots in order to facilitate real-time calculation or 
> incremental calculation, in the case of data lake is becoming more and more 
> of a trend, Real-time computation/incremental computation and snapshot 
> management are very much needed new features of kudu.



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

Reply via email to