Base type used to identify and configure the FailureDetector.
Indicated to use the com.twitter.finagle.liveness.MockFailureDetector when creating a new detector
Helper class for configuring a FailureDetector within a com.twitter.finagle.Stackable client
Indicated to use the com.twitter.finagle.liveness.ThresholdFailureDetector configured with these values when creating a new detector.
Indicated to use the com.twitter.finagle.liveness.ThresholdFailureDetector configured with these values when creating a new detector.
The default windowSize
and threshold
are chosen from examining a
representative ping distribution in a Twitter data center. With long tail
distribution, we want a reasonably large window size to capture long RTTs
in the history. A small threshold makes the detection sensitive to potential
failures. There can be a low rate of false positive, which is fine in most
production cases with cluster redundancy.
closeTimeout
allows a session to be closed when a ping response times out
after 4 seconds. This allows sessions to be reestablished when there may be
a networking issue, so that it can choose an alternative networking path instead.
The default 4 seconds is pretty conservative regarding normal ping RTT.
Default config type which tells the FailureDetector to extract config values from the sessionFailureDetector GlobalFlag.
Indicated to use the com.twitter.finagle.liveness.NullFailureDetector when creating a new detector
Companion object capable of creating a FailureDetector based on parameterized config.