You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@impala.apache.org by "Matthew Jacobs (JIRA)" <ji...@apache.org> on 2017/04/04 23:34:41 UTC

[jira] [Created] (IMPALA-5167) Query should share a single Kudu client

Matthew Jacobs created IMPALA-5167:
--------------------------------------

             Summary: Query should share a single Kudu client
                 Key: IMPALA-5167
                 URL: https://issues.apache.org/jira/browse/IMPALA-5167
             Project: IMPALA
          Issue Type: Improvement
          Components: Backend
    Affects Versions: Impala 2.8.0
            Reporter: Matthew Jacobs


Creating Kudu clients is very expensive as each will fetch metadata from the Kudu master. Eventually we want to create and always reuse a single Kudu client, but this will be tricky as we will have to invalidate its metadata, which will be tricky (specifically knowing when to do so). For now, we should at least re-use a single Kudu client for the duration of a query. We do have the a Kudu client available for re-use on the QueryExecState, but it's not used across the entire BE yet, and we should also try to share the instance with the FE (e.g. wrapping the functionality we need with jni).



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)