As of the upcoming Linkerd 1.2.0 release, we are deprecating the
We’ve been considering removing the
statsd telemeter because it doesn’t work the way most people expect, which can cause big surprises in production. Specifically, the way that the
statsd telemeter samples counter increments means that you can miss out on very important information in your runtime metrics (like failures). To avoid missing important events, the sample rate on the telemeter can be increased, but this will lead to increased latency.
We’ve migrated most of our other datadog/statsd users onto the influx telemeter, which is much less lossy. These users tend to use telegraf to export higher fidelity stats to datadog (https://github.com/influxdata/telegraf/tree/release-1.3/plugins/outputs/datadog).