[ 
https://issues.apache.org/jira/browse/DISPATCH-1547?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Ken Giusti resolved DISPATCH-1547.
----------------------------------
    Fix Version/s:     (was: 1.11.0)
       Resolution: Not A Bug

False alarm!

This is actually expected behavior.  When a remote router's connection drops 
the routing algorithm holds off flushing the router's associated mobile address 
for "remoteLsMaxAgeSeconds" (from the router configuration entity).  This 
allows the routing table to "ride out" a momentary network fault.  Should the 
router re-establish its connection within remoteLsMaxAgeSeconds (and the 
router's route table hasn't changed) then there is no disruption to the global 
routing database.

If the router does not re-establish its connection within remoteLsMaxAgeSeconds 
the router is considered down and the associated mobile addresses are flushed.

By default the remoteLsMaxAgeSeconds is 60, which matches the actual behavior 
observed.

Note:  if the router does reconnect within remoteLsMaxAgeSeconds, but the 
router's route table has changed (the route table sequence number does not 
match the last observed sequence number for that router) then the routing 
database will be synchronized.

In any case I've added new ctest tests to verify the correct behavior.

 

 

> Route table flush on loss of router takes too long
> --------------------------------------------------
>
>                 Key: DISPATCH-1547
>                 URL: https://issues.apache.org/jira/browse/DISPATCH-1547
>             Project: Qpid Dispatch
>          Issue Type: Bug
>          Components: Routing Engine
>    Affects Versions: 1.10.0
>            Reporter: Ken Giusti
>            Assignee: Ken Giusti
>            Priority: Major
>
> Create a router network of 3 routers connected linearly:
>  
>   RouterA <—> RouterB <—> RouterC
>  
> Add subscribers to RouterA.  Wait until subscribers mobile addresses show up 
> on RouterC.  Then kill RouterA.
> Expect: the mobile addresses on RouterC should be flush relatively quickly
> Action: the mobile addresses on RouterC remain until the link is marked as 
> stale (60 seconds in the default config)



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

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@qpid.apache.org
For additional commands, e-mail: dev-h...@qpid.apache.org

Reply via email to