Serilog.Core.Sinks.Batching.FailureAwareBatchScheduler
Manages reconnection period and transient fault response for BatchingSink.
During normal operation an object of this type will simply echo the configured batch transmission
period. When availability fluctuates, the class tracks the number of failed attempts, each time
increasing the interval before reconnection is attempted (up to a set maximum) and at predefined
points indicating that either the current batch, or entire waiting queue, should be dropped. This
Serves two purposes - first, a loaded receiver may need a temporary reduction in traffic while coming
back online. Second, the sender needs to account for both bad batches (the first fault response) and
also overproduction (the second, queue-dropping response). In combination these should provide a
reasonable delivery effort but ultimately protect the sender from memory exhaustion.
namespace Serilog.Core.Sinks.Batching
{
internal class FailureAwareBatchScheduler
{
public TimeSpan NextInterval { get; }
public FailureAwareBatchScheduler(TimeSpan bufferingTimeLimit, TimeSpan retryTimeLimit);
public void MarkSuccess();
public void MarkFailure(out bool shouldDropBatch, out bool shouldDropQueue);
}
}