@rdhabalia I believe that having z-node created and deleted in very different 
parts of the code makes it very hard to figure out how things work. Having the 
entire logic for the lock and its cleanup within the same class encapsulate the 
complexity there and anyone reading that class can easily understand the 
behavior.

Regarding `NoopLoadManager`: the standalone currently works with 
`ModularLoadManager`, but `ModularLoadManager` was really not designed for 
standalone. In fact, standalone mode doesn't need a load manager because there 
are no other brokers. All the features provided by a load manager are not used 
by standalone (and sometimes they complicate it): load-balancing, getting load 
reports, tracking rates of bundles, traffic shedding and so on.  All this is, 
in the best case, it's overhead to the functioning of Pulsar standalone. The 
`NoopLoadManager` short circuits all these features. The only piece of code it 
contains is the z-node registration, since it's needed in lookup logic. 



[ Full content available at: 
https://github.com/apache/incubator-pulsar/pull/2487 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to