package org.infinispan.context;
import java.io.IOException;
import java.io.ObjectInput;
import java.io.ObjectOutput;
import java.util.Collections;
import java.util.EnumSet;
import java.util.Set;
import org.infinispan.AdvancedCache;
import org.infinispan.Cache;
import org.infinispan.commons.marshall.AbstractExternalizer;
import org.infinispan.commons.marshall.MarshallUtil;
import org.infinispan.context.impl.FlagBitSets;
import org.infinispan.lifecycle.ComponentStatus;
import org.infinispan.marshall.core.Ids;
/**
* Available flags, which may be set on a per-invocation basis. These are
* provided using the {@link AdvancedCache} interface, using some of the
* overloaded methods that allow passing in of a variable number of Flags.
*
* When making modifications to these enum, do not change the order of
* enumerations, so always append any new enumerations after the last one.
* Finally, enumerations should not be removed.
*
* @author Manik Surtani
* @author Galder Zamarreño
* @since 4.0
*/
public enum Flag {
/**
* Overrides the {@link org.infinispan.configuration.cache.LockingConfiguration#lockAcquisitionTimeout(long)} configuration setting by ensuring lock
* managers use a 0-millisecond lock acquisition timeout. Useful if you only want to acquire a lock on an entry
* <i>if and only if</i> the lock is uncontended.
*/
ZERO_LOCK_ACQUISITION_TIMEOUT,
/**
* Forces LOCAL mode operation, even if the cache is configured to use a clustered mode like replication,
* invalidation or distribution. Applying this flag will suppress any RPC messages otherwise associated with this
* invocation. Write operations mat not acquire the entry locks. In distributed mode, the modifications performed
* during an operation in a non-owner node are going to L1, if it is enabled, otherwise the operation is a no-op in
* that node.
*/
CACHE_MODE_LOCAL,
/**
* Bypasses lock acquisition for this invocation altogether. A potentially dangerous flag, as it can lead to
* inconsistent data: a Lock is needed to make sure the same value is written to each node replica; a lock
* is also needed to guarantee that several writes on the same key are not applied out of order to an async CacheLoader
* storage engine.
* So this flag is useful only as an optimization when the same key is written once and never again, or as
* an unsafe optimisation if the period between writes on the same key is large enough to make a race condition
* never happen in practice. If this is unclear, avoid it.
*/
SKIP_LOCKING,
/**
* Forces a write lock, even if the invocation is a read operation. Useful when reading an entry to later update it
* within the same transaction, and is analogous in behavior and use case to a <tt>select ... for update ... </tt>
* SQL statement.
*/
FORCE_WRITE_LOCK,
/**
* Skips checking whether a cache is in a receptive state, i.e. is {@link ComponentStatus#RUNNING}. May break
* operation in weird ways!
*
* @deprecated This flag is no longer in use.
*/
@Deprecated
SKIP_CACHE_STATUS_CHECK,
/**
* Forces asynchronous network calls where possible, even if otherwise configured to use synchronous network calls.
* Only applicable to non-local, clustered caches.
*/
FORCE_ASYNCHRONOUS,
/**
* Forces synchronous network calls where possible, even if otherwise configured to use asynchronous network calls.
* Only applicable to non-local, clustered caches.
*/
FORCE_SYNCHRONOUS,
/**
* Skips storing an entry to any configured {@link org.infinispan.persistence.spi.CacheLoader}s.
*/
SKIP_CACHE_STORE,
/**
* Skips loading an entry from any configured {@link org.infinispan.persistence.spi.CacheLoader}s.
* Useful for example to perform a {@link Cache#put(Object, Object)} operation while not interested
* in the return value (i.e. the previous value of the key).
* <br>
* Note that the loader will be skipped even if that changes the meaning of the operation, e.g. for
* conditional or {@link org.infinispan.atomic.DeltaAware} write operations. If that is not intended,
* you should use {@link #IGNORE_RETURN_VALUES} instead.
*/
SKIP_CACHE_LOAD,
/**
* <p>Swallows any exceptions, logging them instead at a low log level. Will prevent a failing operation from
* affecting any ongoing JTA transactions as well.</p>
* <p>This Flag will not be replicated to remote nodes, but it will still protect the invoker from remote exceptions.</p>
* <p>When using this flag with Optimistic caches, lock acquisition happen in the prepare phase at which
* point this flag will be ignored in order to ensure that Infinispan reports the correct exception
* back to the transaction manager. This is done for safety reasons to avoid inconsistent cache contents.</p>
*/
FAIL_SILENTLY,
/**
* When used with <b>distributed</b> cache mode, will prevent retrieving a remote value either when
* executing a {@link Cache#get(Object)} or {@link Cache#containsKey(Object)},
* or to return the overwritten value for {@link Cache#put(Object, Object)} or {@link Cache#remove(Object)}.
* This would render return values for most operations unusable, in exchange for the performance gains of
* reducing remote calls.
* <br>
* Note that the remote lookup will be skipped even if that changes the meaning of the operation, e.g. for
* conditional or {@link org.infinispan.atomic.DeltaAware} write operations. If that is not intended,
* you should use {@link #IGNORE_RETURN_VALUES} instead.
*/
SKIP_REMOTE_LOOKUP,
/**
* Used by the Query module only, it will prevent the indexes to be updated as a result of the current operations.
*/
SKIP_INDEXING,
/**
* Flags the invocation as a {@link Cache#putForExternalRead(Object, Object)}
* call, as opposed to a regular {@link Cache#put(Object, Object)}. This
* flag was created purely for internal Infinispan usage, and should not be
* used by clients calling into Infinispan.
*/
PUT_FOR_EXTERNAL_READ,
/**
* Flags the invocation as a put operation done internally by the state transfer.
* This flag was created purely for internal Infinispan usage, and should not be
* used by clients calling into Infinispan.
*
* Note for internal users: PUT_FOR_STATE_TRANSFER only applies to state transfer-specific actions,
* in order to avoid loading the previous value one should add the IGNORE_RETURN_VALUES flag explicitly.
*/
PUT_FOR_STATE_TRANSFER,
/**
* Flags the invocation as a put operation done internally by the cross-site state transfer. This flag was created
* purely for internal Infinispan usage, and should not be used by clients calling into Infinispan.
*/
PUT_FOR_X_SITE_STATE_TRANSFER,
/**
* If this flag is enabled, if a cache store is shared, then storage to the store is skipped.
*/
SKIP_SHARED_CACHE_STORE,
/**
* This flag has only effect when it's used before calling {@link
* org.infinispan.Cache#stop()} and its effect is that apart from stopping
* the cache, it removes all of its content from both memory and any backing
* cache store.
*
* @deprecated No longer in use.
*/
@Deprecated
REMOVE_DATA_ON_STOP,
/**
* Used by the DistLockingInterceptor to commit the change no matter what (if the flag is set). This is used when
* a node A pushes state to another node B and A doesn't want B to check if the state really belongs to it
*/
SKIP_OWNERSHIP_CHECK,
/**
* Signals when a particular cache write operation is writing a delta of
* the object, rather than the full object. This can be useful in order to
* make decisions such as whether the cache store needs checking to see if
* the previous value needs to be loaded and merged.
*/
DELTA_WRITE,
/**
* Signals that a write operation's return value will be ignored, so reading
* the existing value from a store or from a remote node is not necessary.
*
* Typical operations whose return value might be ignored include
* {@link java.util.Map#put(Object, Object)} whose return value indicates
* previous value. So, a user might decide to the put something in the
* cache but might not be interested in the return value.
*
* This flag is ignored for operations that need the existing value to execute
* correctly, e.g. {@link Cache#get(Object)},
* conditional remove ({@link Cache#remove(Object, Object)}),
* and replace with an expected value ({@link Cache#replace(Object, Object, Object)}).
*
* That means it is safe to use {@code IGNORE_RETURN_VALUES} for all the operations on a cache,
* unlike {@link Flag#SKIP_REMOTE_LOOKUP} and {@link Flag#SKIP_CACHE_LOAD}.
*/
IGNORE_RETURN_VALUES,
/**
* If cross-site replication is enabled, this would skip the replication to any remote site.
*/
SKIP_XSITE_BACKUP,
/**
* Using a synchronous cache (whether replicated or distributed) provides
* the cache caller guarantees that data has been sent to other cluster
* nodes correctly and has been applied successfully. At the network
* level, message delivery acknowledgement protocols are used to provide
* these guarantees.
*
* In order to increase performance and achieve better throughput, it's
* common to use negative acknowledgment protocols to confirm the
* delivery of messages. The problem with these protocols is that if the
* last message is lost, it can be difficult to recover it because a new
* message would need to be sent to find the gap.
*
* Some cache use cases might involve storing an entry in a synchronously
* replicated or distributed cache, and if that store operation fails, the
* application fails to start. One such example is the Infinispan Hot Rod
* server. When it starts, the first thing it does is add its endpoint
* information to a cache which is used to notify clients of topology
* changes. If this operation fails, the server cannot start because
* topologies won't include this node, and no more cache operations
* are attempted.
*
* So, in exceptional use cases such as this, a cluster wide cache update
* should be positively acknowledged by the other nodes, so that if the
* data is lost, it can be retransmitted immediately without the need
* to wait for an extra cluster wide operation to detect the lost message.
* The way to force a particular cache operation to be positively
* acknowledged is to send this flag.
*
* Note that this is flag is <b>EXPERIMENTAL</b> and so there is a high
* probability that it will be removed in future Infinispan versions.
*/
GUARANTEED_DELIVERY,
/**
* This flag skips listener notifications as a result of a cache operation.
* For example, if this flag is passed as a result of a {@link Cache#get(Object)}
* call, no callbacks will be made on listeners annotated with
* {@link org.infinispan.notifications.cachelistener.annotation.CacheEntryVisited}.
*/
SKIP_LISTENER_NOTIFICATION,
/**
* This flag skips statistics updates as a result of a cache operation.
* For example, if this flag is passed as a result of a {@link Cache#get(Object)}
* call, no cache hits or cache miss counters will be updated.
*/
SKIP_STATISTICS,
/**
* Flag to identify cache operations coming from the Hot Rod server.
*/
OPERATION_HOTROD,
/**
* Flag to identify cache operations coming from the Memcached server.
*/
OPERATION_MEMCACHED,
/**
* Any time a new indexed entry is inserted, a delete statement is issued on the indexes
* to remove previous values. This delete statement is executed even if there is no known
* entry having the same key. Enable this flag when you know for sure there is no existing
* entry in the index for improved performance.
* For example, this is useful for speeding up import of new data in an empty cache
* having an empty index.
*/
SKIP_INDEX_CLEANUP,
/**
* If a write operation encounters a retry due to a topology change this flag should be used to tell the new owner
* that such a thing happened. This flag was created purely for internal Infinispan usage, and should not be
* used by clients calling into Infinispan.
*/
COMMAND_RETRY,
/**
* Flag to identity that data is being written as part of a Rolling Upgrade.
*/
ROLLING_UPGRADE
;
/**
* Creates a copy of a Flag BitSet removing instances of FAIL_SILENTLY.
*
* @deprecated Since 9.0
*/
@Deprecated
public static long copyWithoutRemotableFlags(long flagsBitSet) {
return FlagBitSets.copyWithoutRemotableFlags(flagsBitSet);
}
/**
* Creates a copy of a Flag set removing instances of FAIL_SILENTLY.
*
* @deprecated Since 9.0
*/
@Deprecated
public static Set<Flag> copyWithoutRemotableFlags(Set<Flag> flags) {
//FAIL_SILENTLY should not be sent to remote nodes
if (flags != null && flags.contains(Flag.FAIL_SILENTLY)) {
EnumSet<Flag> copy = EnumSet.copyOf(flags);
copy.remove(Flag.FAIL_SILENTLY);
if (copy.isEmpty()) {
return Collections.emptySet();
}
else {
return copy;
}
} else {
return flags;
}
}
private static final Flag[] CACHED_VALUES = values();
private static Flag valueOf(int ordinal) {
return CACHED_VALUES[ordinal];
}
public static class Externalizer extends AbstractExternalizer<Flag> {
@Override
public Integer getId() {
return Ids.FLAG;
}
@Override
public Set<Class<? extends Flag>> getTypeClasses() {
return Collections.singleton(Flag.class);
}
@Override
public void writeObject(ObjectOutput output, Flag flag) throws IOException {
MarshallUtil.marshallEnum(flag, output);
}
@Override
public Flag readObject(ObjectInput input) throws IOException, ClassNotFoundException {
return MarshallUtil.unmarshallEnum(input, Flag::valueOf);
}
}
}