CloudSQL metrics' average value does not match : Per query execution times

[quynhnxq]

Hello. I using Stackdriver Exporter version v0.13.0

• Prometheus scrape_interval: 1m
• Stack driver metrics interval: 1m
• stackdriver metrics: cloudsql.googleapis.com/database/postgresql/insights/perquery/execution_time
• Grafana unit for value: microsecond

In grafana, I used this query to get average time per query:

avg_over_time(stackdriver_cloudsql_instance_database_cloudsql_googleapis_com_database_postgresql_insights_perquery_execution_time{resource_id="$resource_id", database="$database", user="$user", client_addr="$client_addr"}[1h])

But the result on GCP Query Insight is different from the value from the above query. Anyone have solution for this?

Thanks all !

[owenhaynes]

Would say on top of this stackdriver_cloudsql_instance_database_cloudsql_googleapis_com_database_postgresql_insights_* do not work very well when monitoring.aggregate-deltas is enabled. Things like stackdriver_cloudsql_instance_database_cloudsql_googleapis_com_database_postgresql_insights_perquery_row_count do not work as expected.