Hi Devs,

For WSO2 MB, there are situations where in large scale deployments
with hundreds of subscribers connected, network interruptions to
subscribers are sometimes not propagated to Andes library to close them.

This results scale subscriptions hanging on the system, where WSO2 MB
'think' subscriber is active and try to send messages to it. It finds that
it is closed when it try to publish it to the wire and re-queue.  This
re-queing happens rapidly slowing down other queues/topics in broker
infrastructure.

This is not recovered until restart for now.

To overcome, as a workaround I would suggest one of below approaches.

1. When code finds it cannot publish messages as underlying connection is
closed, remove the subscription entry from server.
2. During heartbeat, if heartbeat did not respond, close the broker
3. Improve UI to remove a subscription entry forcefully.

WDYT? No:1 looks promising.

Thanks

-- 
*Hasitha Abeykoon*
Associate Technical Lead; WSO2, Inc.; http://wso2.com
*cell:* *+94 719363063*
*blog: **abeykoon.blogspot.com* <http://abeykoon.blogspot.com>
_______________________________________________
Architecture mailing list
[email protected]
https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture

Reply via email to