WARNING: Version 5.x has passed its EOL date.
This documentation is no longer being maintained and may be removed. If you are running this version, we strongly advise you to upgrade. For the latest information, see the current release documentation.
Not typically something you’ll have to pass to the client but all calls to
in the client have been abstracted behind an
This allows us to unit test timeouts and cluster failover without being bound to wall clock
time as calculated by using
var dateTimeProvider = DateTimeProvider.Default; dateTimeProvider.Now().Should().BeCloseTo(DateTime.UtcNow);
As you can see, dates are always returned in UTC from the default implementation.
Another responsibility of this interface is to calculate the time a node has to be taken out of rotation based on the number of attempts to revive it. For very advanced use cases, this might be something of interest to provide a custom implementation for.
The default timeout calculation is
min(timeout * 2 ^ (attempts * 0.5 -1), maxTimeout)
where the default values for
var timeout = TimeSpan.FromMinutes(1); var maxTimeout = TimeSpan.FromMinutes(30);
Plotting these defaults looks as follows
The goal here is that whenever a node is resurrected and is found to still be offline, we send it back to the doghouse for an ever increasingly long period, until we hit a bounded maximum.
var dateTimeProvider = DateTimeProvider.Default; var timeouts = Enumerable.Range(0, 30) .Select(attempt => dateTimeProvider.DeadTime(attempt, timeout, maxTimeout)) .ToList(); foreach (var increasedTimeout in timeouts.Take(10)) increasedTimeout.Should().BeWithin(maxTimeout);