Open mkuratczyk opened 1 week ago
@mkuratczyk I have added the new suite to Bazel.
We can, even now, but I wouldn't rush with actually enforcing anything. The amount of code needed to support the old name is tiny and there can be applications that have that name in the codebase and we all know that for some users, application code changes are very hard (this particular setting is probably much more common in policies, that are easier to change, but still).
We can, even now, but I wouldn't rush with actually enforcing anything.
No, we can't remove support for queue-master-locator
right now because - as you wrote - applications use queue-master-locator
.
The question is how can we eventually remove support for queue-master-locator
in RabbitMQ (e.g. version 5.0)? The deprecated feature subsystem enables "a way to declare deprecated features in the code, not only in our communication.".
Users might not read the documentation and might not even understand that queue-master-locator
is deprecated.
As explained in https://github.com/rabbitmq/rabbitmq-server/pull/7390#issue-1595515010 deprecation of queue-master-locator
would go through the following lifecycle:
We already use this deprecated feature subsystem for various deprecated features.
It might be a good idea to also "properly" deprecate queue-master-locator
using the deprecated feature subsystem.
As I said on Slack, I also favor the use of deprecated features because it increases the chance that the information reaches the user who rely on that feature. It also allows them to test RabbitMQ as if the feature was gone.
Corresponding docs PR: https://github.com/rabbitmq/rabbitmq-website/pull/1961
queue-leader-locator
queue-leader-locator
takes precedence overqueue-master-locator
if both are usedclient-local
(default) orbalanced
min-masters
,random
,least-leaders
) are mapped tobalanced