You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@calcite.apache.org by "Julian Hyde (JIRA)" <ji...@apache.org> on 2016/07/15 00:35:20 UTC

[jira] [Commented] (CALCITE-1316) Better control over retried operations in Avatica client

    [ https://issues.apache.org/jira/browse/CALCITE-1316?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15378666#comment-15378666 ] 

Julian Hyde commented on CALCITE-1316:
--------------------------------------

The optimal number of knobs is zero. Let's have justification that the knob needs to be twiddled before adding one. I suspect the the only "users" likely to twiddle knobs are you and me, and we might as well change the source code.

> Better control over retried operations in Avatica client
> --------------------------------------------------------
>
>                 Key: CALCITE-1316
>                 URL: https://issues.apache.org/jira/browse/CALCITE-1316
>             Project: Calcite
>          Issue Type: Improvement
>          Components: avatica
>            Reporter: Josh Elser
>            Assignee: Josh Elser
>             Fix For: avatica-1.9.0
>
>
> We have at least two places in the Avatica client now where we will try to re-issue "RPCs" in the attempt to work seamlessly with load-balanced servers.
> Between these two places, we have finite retries, infinite retries and no standardized back-off strategies. We should try to centralize this into one place and make sure that users can override the logic, heaven forbid they come up with some situation where it's necessary.
> Need to investigate the retry-loops we have in the Avatica client now, categorize the loops and come up with a minimal set of knobs to configure the retries, expose those knobs in the JDBC URL string options, and update the documentation.



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