[jira] [Commented] (GEODE-1748) Client should recover instantiators if all servers have been bounced

["Bruce Schuchardt (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/GEODE-1748?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15416078#comment-15416078 ] 

Bruce Schuchardt commented on GEODE-1748:
-----------------------------------------

PdxRegistryRecoveryListener also has this same TODO comment.  It was probably copied from the other class.

> Client should recover instantiators if all servers have been bounced
> --------------------------------------------------------------------
>
>                 Key: GEODE-1748
>                 URL: https://issues.apache.org/jira/browse/GEODE-1748
>             Project: Geode
>          Issue Type: Bug
>          Components: client/server
>            Reporter: Bruce Schuchardt
>
> This TODO in InstantiatorRecoveryListener needs to be addressed:
> {noformat}
>  * TODO - There is a window in which all of the servers could crash and come
>  * back up and we would connect to a new server before realizing that all the
>  * servers crashed. To fix this, we would need to get some kind of birthdate of
>  * the server ds we connect and use that to decide if we need to recover
>  * instantiators. As it is, the window is not very large.
> {noformat}
> I think this means that the Endpoints monitored by this class need to have a timestamp for the creation of the servers' distributed system, which would be a change to the servers' membership view (add a creation timestamp) and a change to the client/server handshake so that this info can be sent to the client.  The listener would then have to check the existing endpoints when a new one is added and make sure they aren't from the old system.



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