I am using Spring Data for Apache Geode for a Spring Boot app that uses a remote Apache Geode server. To set up local Regions I am using @Configuration
class like:
@Configuration
public class GeodeConfig {
public static Region<String, String> myRegion;
@Bean("setupMyRegion")
public Region<String, String> setupMyRegion(GemFireCache cache) {
Region<String, String> r = cache.getRegion("myRegion");
myRegion = r;
return r;
}
}
This enables me to put and get data with the Region.
When I register interest in subscribing to a key, the subscription is setup with scope=LOCAL like this:
org.apach.geode.inter.cache.LocalRegion 4439 logKeys: org.apache.geode.internal.cache.LocalRegion[path='/myRegion';scope=LOCAL';dataPolicy=NORMAL; concurrencyChecksEnabled] refreshEntriesFromServerKeys count=1 policy=KEYS_VALUES
and I guess that I want to set it up as a CACHING_PROXY
so that later, on subscribing to events in the Region, the local subscription will see the server Region events.
I've set subscriptionEnabled=TRUE
for the @ClientCacheApplication
and I am trying to set
clientRegionShortcut = ClientRegionShortcut.CACHING_PROXY
I don't know where to do this. I've tried the @EnableClusterDefinedRegions
annotation but the Region is still being set to local.
If I try to create a Region Factory from the GemFireCache
it gives me an error that the method createClientRegionFactory(ClientRegionShortcut)
is "undefined" for the type GemFireCache
.
THANKS
Generally, it is bad practice to use the Apache Geode API directly to configure beans in a Spring context. For example...
@Bean("MyRegion")
public Region<String, Object> myRegion(GemFireCache cache) {
return cache.getRegion("myRegion");
}
Or using:
@Bean("MyRegion")
public Region<String, Object> myRegion(ClientCache clientCache) {
return clientCache.createClientRegionFactory(ClientRegionShortcut.CACHING_PROXY)
.create("MyRegion");
}
Both approaches above are not recommended!
Spring Data for Apache Geode (SDG) contains specific Spring FactoryBeans
to do the same thing. For instance, the SDG ClientRegionFactoryBean
.
You would use the ClientRegionFactoryBean
in your Spring application configuration as follows:
@Bean("MyRegion")
ClientRegionFactoryBean<String, Object> myRegion(ClientCache clientCache) {
ClientRegionFactoryBean<String, Object> myRegion =
new ClientRegionFactoryBean<>();
myRegion.setCache(clientCache);
myRegion.setShortcut(ClientRegionShortcut.CACHING_PROXY);
return myRegion;
}
There are SDG FactoryBeans
for different types of Geode objects, such a [client] Regions
, Indexes
, DiskStores
, GatewayReceivers
, GatewaySenders
, etc.
These FactoryBeans
are essential in ensuring the the "lifecycle" of the Geode objects (e.g. Region(s)
) are properly synced to the lifecycle of the Spring container. By forgoing the use of SDG's FactoryBeans
you are effectively ignoring the lifecycle coordination provided by the Spring container and custom (framework) FactoryBeans
, and are therefore responsible for the lifecycle of the Geode objects themselves when using Geode's API direct in Spring config. This is one of the primary reasons SDG's exists in the first place.
A "client" Region
must be a *PROXY
(e.g. PROXY
or CACHING_PROXY
) in order to enable subscriptions and receive interesting events from servers for that Region. This is because the Geode cluster (servers) maintain a subscription queue for clients when they connect, which contain events that the client has expressed interests in receiving.
For *PROXY
client Regions
, this also means that the client Region
must have a corresponding server-side, peer cache Region
by the same name.
You can simplify Region creation in general, and client Region creation in particular, by using SDG's Annotation-based configuration model, and specifically, @EnableEntityDefinedRegions
or @EnableCachingDefinedRegions
.
For instance:
@ClientCacheApplication(subscriptionsEnabled = true)
@EnableEntityDefinedRegions(basePackageClasses = SomeEntity.class,
clientRegionShortcut = ClientRegionShortcut.CACHING_PROXY)
class GeodeConfiguration {
...
}
Of course, this assumes you have defined application domain model entity classes, such as:
package example.app.model;
import ...;
@Region("MyRegion")
class SomeEntity {
@Id
private Long id;
...
}
The ClientRegionFactoryBean
would be typed accordingly: ClientRegionFactoryBean<Long, SomeEntity>
.
See the SDG Annotation-based configuration model documentation for more details, and specifically Annotation configuration covering Region definitions.
The @EnableClusterDefinedRegions
annotation defines client-side Regions
for all existing server-side (cluster) Regions
to which the client connects.
For example, if you start a cluster in Gfsh as follows:
gfsh>start locator --name=MyLocator ...
gfsh>start server --name=MyServer ...
gfsh>create region --name=Example --type=PARTITION
gfsh>create region --name=AnotherExample --type=REPLICATE
And then connect a Spring Boot, Apache Geode ClientCache
application to this cluster by using the following Geode configuration:
@ClientCacheApplication
@EnableClusterDefinedRegions(clientRegionShortcut =
ClientRegionShortcut.CACHING_PROXY)
class GeodeConfiguration {
...
}
Then the Spring Boot, Apache Geode ClientCache
application would have 2 Spring configured client Region beans with names "Example" and "AnotherExample" both of type, CACHING_PROXY
.
If there are NO server-side Regions presently in the cluster, then the @EnableClusterDefinedRegions
would not define any client-side Regions.
Anyway, I hope this makes sense and helps!
Cheers!