You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@geode.apache.org by "Alexander Murmann (JIRA)" <ji...@apache.org> on 2018/06/07 20:11:00 UTC
[jira] [Commented] (GEODE-5297) When command tries to execute a
function on a group of members, if one member's function execution failed,
we should consider the command failed.
[ https://issues.apache.org/jira/browse/GEODE-5297?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16505191#comment-16505191 ]
Alexander Murmann commented on GEODE-5297:
------------------------------------------
While this sound like a good change, it seems to me like this cannot happen before Geode 2.0, since it would be a breaking API change.
> When command tries to execute a function on a group of members, if one member's function execution failed, we should consider the command failed.
> -------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: GEODE-5297
> URL: https://issues.apache.org/jira/browse/GEODE-5297
> Project: Geode
> Issue Type: Bug
> Reporter: Jinmei Liao
> Priority: Major
> Labels: geode2
>
> Currently, when we create the command result using ResultModel.createMemberStatusResult from a list of CliFunctionResult, we would set the the status of command result to be successful if there is at least one successful result. This was originally because some function execution will fail for some legitimate reasons like the region is already create/destroyed on this member. Now that since we introduced "ignorable" state in CliFunctionResult and we can ignore the "ignorable" failure based on the commands "if-exist" or "if-not-exist" flags, I suspect we can make the shift from using "at least one success to determine command success" to using "at least one failure to determine command failure".
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)