StatsD repeater backends cause data lose

[rrfeng]

We have a statsd run on UDP and InfluxDB backend worked fine, until the day before yesterday when we add a repeater backend to the statsd instance.
Repeater send data to another statsd and there is a backend write by ourselves, for some extra analysis.
At first the two statsd worked fine, but at 20:00 yesterday, the problem came. We use grafana with InfluxDB backend, and the counter shows only 50/s which should be aoubt 100/s (tps of an api). And same at tonight 20:00. (traffic going heavy at the time.)
I remove the repeater backend and restart the statsd, all things going to be ok.
How should I resolve this?

[elifish4]

Hi@rrfeng ,
I have similar problem.
Did you managed to solve this issue?

[rrfeng]

@elifish4
yes, I found out that was because the cpu load. As you know, nodejs only use single core, when the cpu usage being 100%, it start losing data.

I made a trick:
before:

source -> statsd(repeater and store) ->[( influxdb )(statsd -> xxxxx)]

after:

source -> statsd(repeater only) -> [(statsd -> influxdb)(statsd -> xxxx)]