[GEODE-8324] Events are not being replicated from one WAN site to another when there are multiple paths between the sites and the direct one is stopped

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: wan
• Labels: caching-applications pull-request-available
• Assignee: Barrett Oglesby
• Reporter: Barrett Oglesby
• Created: 30/Jun/20 21:03
• Updated: 26/Mar/21 23:23

Description

Scenario
--------
The configuration is a normal star/mesh pattern where each site is connected to each other site.

So, there are 2 paths from site-A to site-B:

• the direct (site-A -> site-B)
• the indirect path (site-A -> site-C -> site-B)

If the direct path (meaning the sender in site-A to site-B) is stopped and a put is done in site-A, site-B doesn't receive the event even though the indirect path is not stopped.

Here is the behavior:

Site-A - dsid=1
Site-B - dsid=2
Site-C - dsid=3

Site-A
------
Since site-A is configured with site-B and site-C, its events will contain recipient dsids for those sites (2 and 3 below).

The event sent to site-C will look like:

AbstractGatewaySender.distribute sender=site-C processing local event newCallbackArg=GatewaySenderEventCallbackArgument [originatingSenderId=1;recipientGatewayReceivers={3, 2}]; region=/data; key=0

Site-C
------
Site-C receives the event:

GatewayReceiverCommand.cmdExecute processing create region=/data; key=0

The sender to site-B drops the event since it thinks it is already a recipient:

AbstractGatewaySender.distribute sender=site-B received event from remote site receivedCallbackArg=GatewaySenderEventCallbackArgument [originatingSenderId=1;recipientGatewayReceivers={3, 2}]; region=/data; key=0; eventId=EventID[id=25 bytes;threadID=0x10030|1;sequenceID=0;bucketId=48]
AbstractGatewaySender.distribute sender=site-B dropping event since it is already a recipient

Issue Links

• links to: GitHub Pull Request #5331