@Stability.Internal
public class LostCleanupDistributed
extends java.lang.Object
The user specifies that lost transactions should be found inside a certain cleanup window of X seconds.
As this algo is distributed and tries to require minimal co-operation between clients (they simply read and write a Client Record doc periodically, there are flaws including:
Two clients can read the same client record at the same time and miss each other's updates A client can read the record just before a new client removes itself
The idea is that these conflicts will sort themselves out on the next iteration, and that in the vast majority of cases each ATR will be checked in X seconds.
Constructor and Description |
---|
LostCleanupDistributed(TransactionConfig config,
ClusterData clusterData,
java.util.function.Supplier<Cleaner> cleanerSupplier) |
Modifier and Type | Method and Description |
---|---|
static java.util.List<java.lang.String> |
atrsToHandle(int indexOfThisClient,
int numActiveClients,
int numAtrs) |
java.lang.String |
clientUuid() |
reactor.core.publisher.Flux<TransactionCleanupAttempt> |
handleATRCleanup(java.lang.String bp,
com.couchbase.client.java.ReactiveCollection atrCollection,
java.lang.String atrId,
ATRStats stats,
java.time.Duration safetyMargin)
Looks at an ATR, finds any expired entries, and cleans them up.
|
void |
stop() |
void |
stopThreads() |
public LostCleanupDistributed(TransactionConfig config, ClusterData clusterData, java.util.function.Supplier<Cleaner> cleanerSupplier)
public java.lang.String clientUuid()
public void stop()
public void stopThreads()
public static java.util.List<java.lang.String> atrsToHandle(int indexOfThisClient, int numActiveClients, int numAtrs)
public reactor.core.publisher.Flux<TransactionCleanupAttempt> handleATRCleanup(java.lang.String bp, com.couchbase.client.java.ReactiveCollection atrCollection, java.lang.String atrId, ATRStats stats, java.time.Duration safetyMargin)