Closed ennru closed 5 years ago
Should I fix all those point in a single commit ?
Split it in several PR if you like to do it in small steps, but one PR is OK, as well.
While reviewing #1096 I noticed a couple of other changes that should be done, to make Geode connector more inline with the others:
ReactiveGeode
to be a collection of factory methods instead of a class that needs to be instantiated. That will mean that cache and close methods will have to move to stages themselves.Sorry for the late reply. But AFAIK having more then one (active) cache per jvm is not that easy (https://geode.apache.org/releases/latest/javadoc/org/apache/geode/cache/client/ClientCacheFactory.html#create--) I will try to investigate further.
Is there a chance that akka extension mechanism should fit to hold the Geode cache (aka connection) ?
We use Akka Extensions in some connectors (reference
as an example, dynamodb
, google-cloud-pub-sub-grpc
, ...) to hold resources that might be shared across instances of Source
s, Flow
s and Sink
s.
However Akka Extension allows to have a resource per ActorSystem
. If that is fine for the Geode cache, then yes, we could use it. But you mentioned that cache is per JVM. Then I think we should leave it in some static field on an uninstantiable class.
impl
package (#1096)@InternalApi
andprivate[connector]
(#1096)docs.javadsl
anddocs.scaladsl
(#1096)