[+] add new stat activity grid report

[pmpetit]

Hello,
One of my collegue Manu G. has created a new report on stat-activity, to match our needs.
I have tested it, i think we should share.

Example

[pashagolub]

Can you please describe this dashboard in a few words?
Thanks in advance!

[pmpetit]

This dashboard represents the number of sessions taht are executing the query, just like oracle does in oem (Oracle Entreprise Mngt).
For example in yellow at 9h55, we can see that 4 sessions were executing
SELECT /* getReception etc... (the second query)

It is a well known dashboard in oracle world.

[pashagolub]

Thanks a lot for clarifications!

[pashagolub]

During testing this grid I got an error from Grafana:

Templating
Failed to upgrade legacy queries Datasource pgwatch-metrics was not found
Templating [dbname]
Error updating options: Datasource pgwatch-metrics was not found

Did you see such messages on a fresh installs with your dashboards?

[pashagolub]

I noticed that only your dashboartd.json contains "datasource": "pgwatch-metrics" string. Can this be the reason?

[pmpetit]

I have a check

[pashagolub]

I'm not sure what am I doing wrong but it's just doesn't work:

Would you please try to reproduce my steps. In the local pgwatch2 repo execute:

$ ./build-docker-postgres.sh

$  docker run -d --restart=unless-stopped --name pw2 -p 3000:3000 -p 8080:
8080 -p 127.0.0.1:5432:5432  -e PW2_TESTDB=true   cybertec/pgwatch2-postgres:latest

Then try to open http://localhost:3000/d/stat-activity/stat-activity

[pmpetit]

Hi,
i have troubleshooted, it is ok now:

• add this metric to default "FULL" template (we were using only the preset-config.yaml file), that's why you could not see it in your test.
• troubleshoot some typo (epoch was missing in the metric)

i have tested with your docker run command line (and pgbench):

i wonder if i should still limit to 100ms:

select /* pgwatch2_generated */
  (extract(epoch from now()) * 1e9)::int8 as epoch_ns,
  s.query as query,
  count(*) as count
from get_stat_activity() s
where s.datname = current_database()
  and s.state = 'active'
  and s.backend_type = 'client backend'
  and s.pid != pg_backend_pid()
  and now() - s.query_start > '100ms'::interval
group by s.query;

or set duration to "10 ms" and filter at dashboard level.

for the moment, i fix it to 100ms.

[pashagolub]

Excellent! Thank you.