Hey there,
I'm using Stackdriver to monitor micro services via submitting different custom metrics. Each micro service locally aggregates metrics and submits a single point for each time series every minute to Stackdriver. But I still do see frequent errors like this for common metrics:
One or more TimeSeries could not be written: One or more points were written more frequently than the maximum sampling period configured for the metric.
For example, there is
custom.googleapis.com/grpc/response/time_ms common metric that each micro service is submitting. Each micro service is submitting each metric with some common labels for this particular container/service like
pod name and
service name. It makes each metric unique across all containers that submit metrics.
So my question is, does Stackdriver actually take into consideration labels when writing time series? It seems Stackdriver only respects a type of a submitted time series but not labels coming with it. In this case a more general question will be, from Stackdriver perspective, how one should submit time series with the same type but different labels (for example when there are many containers for the same service and each container submits metrics to Stackdriver)?
Best,
Fedor