Skip to main content

Snapshots

Snapshots checkpoint the actor's projected state at a known sequence number so that recovery can skip replaying the full event history from the beginning.

When to snapshot​

Without snapshots, recovery replays every event the actor has ever written. For long-lived aggregates this grows unbounded. A snapshot sets a recovery baseline: on startup, the engine loads the latest snapshot and only replays events written after it.

The overhead of taking a snapshot is one store write per N events. The payoff is proportional to how many events you skip on recovery. For actors that write thousands of events before stopping, snapshots are essential.

SnapshotStrategy​

SnapshotStrategy controls when the engine takes a snapshot. Configure it on EventSourcedBehavior:

src/Aggregates/OrderActor.php
<?php

declare(strict_types=1);

use Monadial\Nexus\Persistence\EventSourced\EventSourcedBehavior;
use Monadial\Nexus\Persistence\EventSourced\SnapshotStrategy;

$behavior = EventSourcedBehavior::create($persistenceId, $emptyState, $commandHandler, $eventHandler)
    ->withEventStore($eventStore)
    ->withSnapshotStore($snapshotStore)
    ->withSnapshotStrategy(SnapshotStrategy::everyN(50))
    ->toBehavior();

Three built-in strategies:

FactoryBehaviour
SnapshotStrategy::everyN(int $n)Snapshot after every N-th event (sequence number divisible by N)
SnapshotStrategy::never()Never snapshot — all recovery replays from sequence 0
SnapshotStrategy::predicate(Closure $fn)Custom predicate: fn(state, event, seqNr): bool
note

A snapshot store must be configured via ->withSnapshotStore($store) for snapshots to be taken. Setting a strategy without a store is silently ignored.

What a snapshot stores​

The engine wraps the projected state in a SnapshotEnvelope:

FieldTypeDescription
persistenceIdPersistenceIdIdentity of the actor stream
sequenceNrintSequence number at the time of the snapshot
stateobjectThe serialized projected state
stateTypestringFQCN of the state class
timestampDateTimeImmutableWall-clock time of the snapshot
writerIdUlidIdentity of the ActorSystem that wrote it

The state object is serialized by the store's configured MessageSerializer (default: PhpNativeSerializer).

RetentionPolicy​

After a snapshot is saved, old snapshots and — optionally — the events they cover can be pruned. Configure this via RetentionPolicy:

src/Aggregates/OrderActor.php
<?php

declare(strict_types=1);

use Monadial\Nexus\Persistence\EventSourced\EventSourcedBehavior;
use Monadial\Nexus\Persistence\EventSourced\RetentionPolicy;
use Monadial\Nexus\Persistence\EventSourced\SnapshotStrategy;

$behavior = EventSourcedBehavior::create($persistenceId, $emptyState, $commandHandler, $eventHandler)
    ->withEventStore($eventStore)
    ->withSnapshotStore($snapshotStore)
    ->withSnapshotStrategy(SnapshotStrategy::everyN(50))
    ->withRetention(RetentionPolicy::snapshotAndEvents(keepSnapshots: 3, deleteEventsTo: true))
    ->toBehavior();
FactoryBehaviour
RetentionPolicy::none()Keep all snapshots and all events forever
RetentionPolicy::snapshotAndEvents(int $keep, bool $deleteEventsTo)Keep the latest $keep snapshots; when $deleteEventsTo is true, also delete events up to the oldest retained snapshot's sequence number
Event deletion is permanent

Setting deleteEventsTo: true calls EventStore::deleteUpTo() after each snapshot. Events deleted this way cannot be recovered. Do not enable this if you need to project events into read models.

Recovery acceleration in practice​

With SnapshotStrategy::everyN(50) and a RetentionPolicy that keeps 3 snapshots:

  • An actor with 10,000 events and a snapshot at event 9,950 recovers by loading one snapshot and replaying 50 events.
  • Without snapshots, the same actor replays all 10,000 events.

For actors that write many small events frequently (counters, metrics accumulators), use a smaller N (e.g. everyN(10)). For aggregates that process occasional domain commands, everyN(100) is a reasonable default.

See also​