Currently in `FailbackClusterInvoker` there is a map to store the failed 
invocations, however the size is unlimited.

In a high throughput application, It is possible that the failed invocations 
will fill up the memory and exhaust the memory.

I think it is better to:
* limit the size of the map
* choose the strategy to discard the failed invocations. e.g. FIFO, LIFO, etc.

[ Full content available at: 
https://github.com/apache/incubator-dubbo/issues/2425 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to