You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@kudu.apache.org by "Mike Percy (JIRA)" <ji...@apache.org> on 2017/01/19 19:45:26 UTC

[jira] [Created] (KUDU-1841) Add a client write mode like AUTO_BACKGROUND_FLUSH + data dependencies

Mike Percy created KUDU-1841:
--------------------------------

             Summary: Add a client write mode like AUTO_BACKGROUND_FLUSH + data dependencies
                 Key: KUDU-1841
                 URL: https://issues.apache.org/jira/browse/KUDU-1841
             Project: Kudu
          Issue Type: New Feature
          Components: client
            Reporter: Mike Percy


After considering KUDU-1767, it is clear that we could provide a friendlier client mode for ingest use cases. I think a useful one would be a mode that "feels" like AUTO_BACKGROUND_FLUSH - with only a loose idea of ordering to maximize throughput and no requirement for an explicit flush - but that would also implement a "mutex" for each unique key being written such that writes to a given row always occur in order relative to later writes to the same row, as seen by the client session and the data store.

Such a mode might also be looked at as providing "eventual consistency" because after quiescing writes and reading the latest values it would appear that the data was written in client session order.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)