Marton Elek created HDDS-3315:
---------------------------------
Summary: Use EventQueue for delayed/immediate safe mode rule
notification
Key: HDDS-3315
URL: https://issues.apache.org/jira/browse/HDDS-3315
Project: Hadoop Distributed Data Store
Issue Type: Improvement
Components: SCM
Reporter: Marton Elek
Assignee: Marton Elek
SCM is built from loosely coupled components which communicate with async event
with each other.
Using the same abstraction (EventQueue) has the benefit that we can use the
same visibility / testing tools such as the 'ozone insight' definition (which
makes visible all the messages) or the test handler (which can wait until all
the event queue messages are processed)
During the review of HDDS-3221 it was suggested (by me) to use the EventQueue
instead of the new SafeModeNotification interface.
There was only one counter argument against it:
bq. I personally find the event queue logic hard to follow due to its async
nature (you cannot just follow method calls in the IDE). Its not bad, but more
difficult when you don't yet understand it, while registering some instances to
be notified is easy to follow in an IDE. This is of course a subjective opinion
:)
I respect this opinion, but I think it's better to use one abstraction and a
consistent architecture inside one component (together with all the existing
limitations). The EventQueue is not the only one possible solution, but an
existing one. We can either design and switch to a new one or use the existing
one.
In this patch I would like to show how the previous listener interface can be
replaced by the EventQueue.
It (hopefully) shows that this is not complex, and in fact can help us to
decouple different component from each other
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
