Basically Available Soft statE
- Simple Couchbase Client (Couchbase bucket as distributed / persistent map)
- Manifest comparator : Versioned key for loading refreshed state
Manifest comparison stores a manifest along with each value. The manifest contains the version for the value, if the version has changed, the latest verson of the value will be loaded.
key : manifest [contains version info]
versionedKey : key with version
This allows large immutable datastructures to be stored in as a key/value pair (with separate key/value pairing for version info), and reloaded automatically on change.
Inject the Spring bean created by micro-couchbase ManifestComparator into your own beans. Customise the key used (allows multiple ManifestComparators to be used)
public class ManifestComparatorResource {
private final ManifestComparator<String> comparator;
@Autowired
public ManifestComparatorResource(ManifestComparator comparator) {
this.comparator = comparator.<String>withKey("test-key");
}See micro-events, for Microserver help in managing scheduled jobs.
Create a scheduled job to check the manifest for changes & automatically reload data when stale.
In the example below we run the versioned key cleaner once per day, and check for changes every minute.
@Component
public class DataLoader implements ScheduledJob<Job>{
private final ManifestComparator<String> comparator;
@Autowired
public DataLoader(ManifestComparator comparator) {
this.comparator = comparator.<String>withKey("test-key");
}
@Override
public SystemData<String,String> scheduleAndLog() {
try{
boolean changed = comparator.isOutOfDate();
comparator.load();
return SystemData.<String,String>builder().errors(0).processed(isOutOfDate?1:0).build();
}catch(Exception e){
return SystemData.<String,String>builder().errors(1).processed(0).build();
}
}
}
@Component
public class DataCleaner implements ScheduledJob<Job>{
private final ManifestComparator<String> comparator;
@Autowired
public DataCleaner(ManifestComparator comparator) {
this.comparator = comparator.<String>withKey("test-key");
}
@Override
public SystemData<String,String> scheduleAndLog() {
try{
comparator.cleanAll();
return SystemData.<String,String>builder().errors(0).processed(1).build();
}catch(Exception e){
return SystemData.<String,String>builder().errors(1).processed(0).build();
}
}
}
@Component
public class Schedular{
private final DataCleaner cleaner;
private final DataLoader loader;
public Schedular(DataCleaner cleaner,DataLoader loader){
this.cleaner = cleaner;
this.loader = loader;
}
@Scheduled(cron = "0 1 1 * * ?")
public synchronized void scheduleCleaner(){
cleaner.scheduleAndLog();
}
@Scheduled(cron = "0 * * * * *")
public synchronized void scheduleLoader(){
loader.scheduleAndLog();
}
}Elsewhere a single writer service can write data to the store for all services of a type to load. See micro-mysql or micro-curator for a DistributedLock implementation
e.g.
@Component
public class DataWriter {
private final DistributedLockService lockService;
private final ManifestComparator<String> comparator;
@Autowired
public DataWriter(DistributedLockService lockService,ManifestComparator comparator) {
this.lockService = lockService;
this.comparator = comparator.<String>withKey("test-key");
}
public void write(Supplier<String> data){
if(lockService.tryLock("single-writer-lock-a") {
comparator.saveAndIncrement(data.get());
}
}
}
<dependency>
<groupId>com.aol.microservices</groupId>
<artifactId>micro-couchbase</artifactId>
<version>x.yz</version>
</dependency> compile 'com.aol.microservices:micro-couchbase:x.yz'