vectordotdev
Document that metric sinks aggregate metrics within a batch
Most of the metric sinks aggregate metrics within a batch. That is, incremental metrics are added together and absolute metrics are replaced by the later value. As noted by https://github.com/timberio/vector/pull/7846#pullrequestreview-686985284 this is not adequately described in the documentation for those sinks. This applies to the following sinks:
-
aws_cloudwatch_metrics -
datadog_metrics -
stackdriver_metrics -
influxdb_metrics -
prometheus_exporter -
prometheus_remote_write -
sematext_metrics
I was tripped up by this when naively assuming I could import a backlog of metrics from CSV files to influxdb.
My workaround is specifying a batch.max_events: 1, which is terrible for performance, of course.
I my opinion, this aggregation should not occur in the first place, but that is a different discussion, I guess.
How long before influxdb_metrics can support turning off auto-aggregation?
