[ 
https://issues.apache.org/jira/browse/GEODE-8918?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17310973#comment-17310973
 ] 

ASF subversion and git services commented on GEODE-8918:
--------------------------------------------------------

Commit 73a2120d9c6bca1cd62580414429a40022fe64ca in geode's branch 
refs/heads/feature/GEODE-9081-BufferPool-allocations from Mario Kevo
[ https://gitbox.apache.org/repos/asf?p=geode.git;h=73a2120 ]

GEODE-8918: fix replication in multisite systems (#6008)

* GEODE-8918: fix replication in multisite systems

> Geode replication event forwarding does not honor GW sender state
> -----------------------------------------------------------------
>
>                 Key: GEODE-8918
>                 URL: https://issues.apache.org/jira/browse/GEODE-8918
>             Project: Geode
>          Issue Type: Bug
>          Components: wan
>            Reporter: Mario Kevo
>            Assignee: Mario Kevo
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 1.15.0
>
>
> {color:#172b4d}With 3+ geo-red systems Geode replication has the forwarding 
> feature which means that receiving cluster will forward the event it just got 
> to all clusters it is connected to that have not yet received the 
> event.{color}
> {color:#172b4d}This is possible because the originating cluster is setting 
> metadata in the replication event like this:
> GatewaySenderEventCallbackArgument 
> [originalCallbackArg=null;originatingSenderId=1;recipientGatewayReceivers= 
> {color}{color:#172b4d}{3, 2}{color}{color:#172b4d}]
>  
> Site receiving this event thus knows which is the originating site and which 
> sites should have received this event. All others will have this event 
> forwarded to. All this is legacy Geode behavior.
>  
> However, originating site does not care if GW sender to a destination is 
> stopped or not - only the fact GW sender is *created and attached* to a 
> region is enough. This means if e.g. GW sender from Site1 to Site 3 is 
> stopped (and has been stopped for a while - so this has nothing to do with 
> timing) at the moment an event hits the replication it is only going to be 
> sent to Site 2 *but with the same metadata*_._ Hence Site 2 will not forward 
> to Site 3 (assuming it has a connection to it).{color}



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to