jxblum
commented
1 year ago There are 2 different bits to this equation.
SERVER-SIDE
First, you need to consider what is expected on the server-side, between the peers in the cluster. Each server-side, Spring-configured Locator application (@LocatorApplication) should be configured with the Apache Geode locators property pointing to the other Locator.
For example, on hostOne, running LocatorOne, the Apache Geode locators property (i.e. spring.data.gemfire.locators) or the corresponding @LocatorsApplication annotation, locators attribute, should be set as follows:
@LocatorApplication(name = "LocatorOne", port = 11235, locators="hostTwo[12480]")
class MySpringConfiguredApacheGeodeLocatorApplication {
// ...
}Alternatively, in Spring Boot application.properties for LocatorOne running on hostOne you can set:
# Spring Boot application.properties for LocatorOne running on hostOne
spring.data.gemfire.locators=hostTwo[12480]Subsequently, on hostTwo, LocatorTwo should be configured as follows:
@LocatorApplication(name = "LocatorTwo", port = 12480, locators="hostOne[11235]")
class MySpringConfiguredApacheGeodeLocatorApplication {
// ...
}Or, in Spring Boot application.properties for LocatorTwo running on hostTwo, using:
# Spring Boot application.properties for LocatorTwo running on hostTwo
spring.data.gemfire.locators=hostOne[11235]NOTE: You do not need to configure the Locator port on either the Locator or the client-side if you are using the default port, 10334.
NOTE: Also, the hostnames running the Locator applications must either be the DNS names of your machines or the IP address.
CLIENT-SIDE
Now, for the client-side, you must configure your Apache Geode client Pools with the complete list of Locators in the server-side cluster. For example (using the "DEFAULT" Pool):
@ClientCacheApplication(name = "MyClient" locators = {
@ClientCacheApplication.Locator(host = "hostOne", port=11235),
@ClientCacheApplication.Locator(host = "hostTwo", port=12480)
})
class MySpringConfiguredApacheGeodeClientCacheApplication {
// ...
}Alternatively, and recommended, you can of course, always set the appropriate property:
# Spring Boot application.properties for the Spring-configured, Apache Geode ClientCache application
spring.data.gemfire.pool.locators=hostOne[11235],hostTwo[12480]Of course, you can configure any specifically "named" Pool in either SDG's annotation-based configuration or in Spring Boot application.properties, see here.
Also, you have an option to configure client-side Pool properties (attributes) using SDG's PoolConfigurer (see Javadoc; also see Ref Guide) if you need finer-grained, programmatic control (e.g. conditional) over the Pool configuration, such as hosts and ports.
The same sort of programmatical, Configurer based configuration approach is also applicable on the server-side, e.g. LocatorConfigurer.
Anyway, once the configuration is setup properly, then what Apache Geode (or GemFire) "does" and "shows" is outside of the control of Spring (Data Geode).
For example, if the list clients Gfsh command (here) is not working correctly, then this might be an issue in Apache Geode, and specifically with Gfsh.
Also, the Gfsh list clients command specifically states:
"Display a list of connected clients and the servers to which they connect."
In this case, the Gfsh list clients command might only list "servers", as in specifically CacheServer nodes. I don't remember specifically off the top of my head and you might need to test this!
Technically, cache clients only use the Locators to find servers in the cluster. Once the topology is known, the clients perform data access operations (in)directly (depending on the "single-hop" configuration and Region type, of course (e.g. PARTITION Regions) to the data (server) node with the data of interest.
Anyway, hope this helps for now.
sjoshid
Hi, I have two standalone locators (let's call them L1 and L2) running on two different machines. I have two instances of my app that connect to L1 on startup. But if I bring down L1, I expect them to show up under L2. But that doesnt seem to happen.
Is that how it's suppose to work or Im misunderstanding this?
Env Details
locatorsconfig.How to reproduce?
gfsh. Let's call them L1 and L2 respectively.locatorsconfig.Expectation
None of those expectations are met.
One interesting note, if I start another peer, it shows up in L2 (because L1 is down) but the regions are NOT replicated. So at this point, the last peer (and peers started after) has its own replicated region. So I end up having two disjoint replicated regions which is problematic.