Table db_stats is not updated

[rcapliez]

Hi all !

We have installed pgwatch3 with a PostgreSQL DB deployed on Aiven. It does not fully work. A lot of metrics are working and that's fine but some not ... For example, the dashboard "DB Overview" is empty for "Tuple ins. / upd. / del. statistics (5m avg. rate, log. scale)" or "Shared Buffers hit ratio + Rollback ratio (5m avg.)".

We are using a metrics.yaml file.

We have applied the workaround to avoid to use pg_stat_file but will still get error message where pg_stat_file appears. Where could be the configuration persistent as our pgwatch is running on Kube ... ?

Moreover, the documentation and ChatGPT are mentionning a "pgwatch" schema that we do not find :

\dt pgwatch.*

What are we missing ?

Regards
Raphaël

[pashagolub]

Hello.

There are so many questions...

How do you start pgwatch? Please show the exact command line

[rcapliez]

Hi !
I have understood what happened. I forgot to specify the metrics file in the command line used to launch pgwatch. Pgwatch was using the default metrics. It was not so obvious for me.

The last question that should perhaps need attention is 👍
"Moreover, the documentation and ChatGPT are mentionning a "pgwatch" schema that we do not find :

\dt pgwatch.*"

Is it a mistake from ChatGPT ?
Does / Should this schema exist ?

Best regards

Raphaël

[pashagolub]

To answer your question I need to know how you start pgwatch 🤦

[rcapliez]

This the command line we are using to launch pgwatch :

". /custom/env.sh && /pgwatch/pgwatch --log-level=debug
--metrics=$PW_METRICS_FOLDER --sink=$PW_PG_METRIC_STORE_CONN_STR"

We only see 2 partitions which are "admin" and "subpartitions"

[pashagolub]

ok, now I see.

You use YAML configuration to run pgwatch. And you use Postgres sink to store measurements.

But pgwatch schema is part of Database configuration, e.g.
pgwatch --sources=postgresql://.../pgwatch --sink=$PW_PG_METRIC_STORE_CONN_STR

[rcapliez]

Ok I see.
Thank you for the clarification ;-)
Regards,
Raph

[pashagolub]

You also need to specify --sources=my_monitored_db.yaml in your case.

[rcapliez]

We can close this discussion