[DEPRECATED] Promscale is a unified metric and trace observability backend for Prometheus, Jaeger and OpenTelemetry built on PostgreSQL and TimescaleDB.
In HA mode, Promscale writes from a leader replica and discards from the non-leader. This is true for samples. But, when doing metric-metadata, I found that Promscale was ingesting from non-leader as well. While this is not breaking, it does have some impact on the performance. My guess is we do not go through the writePreprocessors for metric-metadata which can be the cause.
In HA mode, Promscale writes from a leader replica and discards from the non-leader. This is true for samples. But, when doing metric-metadata, I found that Promscale was ingesting from non-leader as well. While this is not breaking, it does have some impact on the performance. My guess is we do not go through the
writePreprocessors
for metric-metadata which can be the cause.