Why for metrics where rows are selected, do you use the dimension in bytes?

eg
pg_stat_database_tup_fetched
Realino says nothing about the instance, if we treat each line as a byte.

Upgrading Postgres results in multiple results being returned for the pg_static{...} query. Switching on 'instant' at least gives you the most recent value of the multiple values.

The panels "Insert data", "Return data", "Fetch data", "Current fetch data" are all using a "size" unit which is wrong as pg_stat_database_tup_* metrics return number of rows...

It looks like some queries in your dashboard should be using rate(). For example, the "Fetch data (SELECT)" chart uses:

pg_stat_database_tup_fetched{datname=~"$datname", instance=~"$instance"}

But pg_state_database_tup_fetched is a counter, and is generally charted via rate():

rate(pg_stat_database_tup_fetched{datname=~"$datname", instance=~"$instance"}[5m])

It looks like the "Insert data", "Update data", and "Return data" charts (at least) have the same issue.

I use one postgres_exporter for several databases, I see duplicate labels in the dashboard, how is it easier to implement an additional filter?

Hi, I couldn't add a "Question" label - but that's the intent of this issue:

1. This dashboard uses postgress-exporter, does it need to be installed on the same machine where my Postgres database is? Or will it work if installed on a different machine and just provide the DB connection details?

I'm asking that mainly for the CPU & Memory usage metrics.. For some reason I see my DB has very low RAM usage, seems wayyy too low IMO (FYI, I installed PG using docker on a k8s pod and postgres-exporter on another pod using docker).

2. How can I understand the metrics around CPU time? Usually, you see CPU usage as a percentage, but here it's CPU time. How can one tell if it's low/mid/high? Maybe you can help transfer it to percentage instead of CPU time?

3. Does it report also the storage of the DB? Meaning how much storage the DB is using. And even better, how much storage it has left on the machine..

4. I see also the "Start Time" is always NA for me.

Is that a problem on my end? Or in the dashboard?

5. More of a pg-exporter question, but any special instructions on how to create the readonly user so everything will work? I did it this way: https://github.com/prometheus-community/postgres_exporter#running-as-non-superuser

Did I miss anything?

Thank you very much!

Getting this error message while trying import this dashboard: 'Invalid dashboard UID in annotation request'

Please kindly help in this regard.

When many pods that expose metrics are deployed it is quite a mess to find the correct one.
I think it has to do with the "Instance" variable in the dashboard. The query filters for all instances that have an "up" metric as far as I can tell. Unfortunately there can be many instances that have a metric with this name.
Is it possible to just use "pg_up" instead of "up"?