[-] fix pgbouncer dashboard

[slardiere]

Fix the database name lookup in the pgbouncer dashboard

• since the metrics database model has changed, the actual database name is inside the JSON field data
• also clean some old and unused sql queries

[coveralls]

Pull Request Test Coverage Report for Build 19141559345

Details

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 71.383%

---

Totals
Change from base Build 19139967054:	0.0%
Covered Lines:	3542
Relevant Lines:	4962

---

💛 - Coveralls

[pashagolub]

Database model isn't change. :) That's something to be done with maintenance routines. Thanks! I'll check!

[pashagolub]

Seems like a promQL to me. Should it be here?

[slardiere]

Probably a remaining query from InfluxDB

[pashagolub]

oh, what an artifact! :)

[pashagolub]

ok, I see it. So the mess is due to terminology. In old days under "dbname" we mean the source from where we fetch metrics. But you want to have exact database on you graph.
Now, with your approach we might have a problem. What if you will have several pgbouncer sources and all of them will have databases with the same names. For example, you have test, stage, prod.
You added 3 pgbouncer sources. If now we will use your approach the data will be mixed.

So my suggestion would be to have 2 drop-downs:

• "Monitored Source" that will filter out per-pgbouncer instance
• "Database" that will filter per-database of per-pgbouncer (or ALL)

Does it sound right to you?

Thanks a lot for your PR!

[slardiere]

Ok, I see, I'll provide that asap

[slardiere]

Hi,

Do you agree with these changes? Or do you need anything else?

best regards,
Sébastien

[0xgouda]

Do you agree with these changes? Or do you need anything else?

We are trying to think of a way to optimize the query so we don't do a full table scan on pgbouncer_stats, which might slow down the dashboards.

[slardiere]

We are trying to think of a way to optimize the query so we don't do a full table scan on pgbouncer_stats, which might slow down the dashboards.

Indeed, thanks,

I just fix the monitoredsource one, but no clue for the database one.

[0xgouda]

I just fix the monitoredsource one, but no clue for the database one.

Thanks!, for the dbname variable, we thought of adding $__timeFilter(time) and making the field update on every interval change, but it turns out where dbname = '$monitoredsource' already utilizes an existing "pgbouncer_stats_dbname_time_idx" btree (dbname, "time") index, so I think it's good enough.

Currently, I am trying to make the dbname multi-value field, and for each new value, new separate panels should appear for it.

[pashagolub]

Thanks @0xgouda.

1. Please include All option into $dbname
2. Make sure all graphs are on the same panel. I don't want repeating, it produces a lot of noise. Here is the example of how it should look like:

[pashagolub]

but it turns out where dbname = '$monitoredsource' already utilizes an existing "pgbouncer_stats_dbname_time_idx" btree (dbname, "time") index, so I think it's good enough.

Well, I cannot confirm that:

pgwatch_metrics=# explain analyze select distinct (data->>'database') as database from pgbouncer_stats where dbname = 'demo_pgbouncer';
                                                                                  QUERY PLAN                                                                                   
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 HashAggregate  (cost=67.47..69.33 rows=149 width=32) (actual time=1.533..1.537 rows=2.00 loops=1)
   Group Key: ((pgbouncer_stats.data ->> 'database'::text))
   Batches: 1  Memory Usage: 32kB
   Buffers: shared hit=7 read=27 dirtied=1
   I/O Timings: shared read=0.831
   ->  Append  (cost=0.00..66.83 rows=255 width=32) (actual time=0.331..1.470 rows=251.00 loops=1)
         Buffers: shared hit=7 read=27 dirtied=1
         I/O Timings: shared read=0.831
         ->  Seq Scan on pgbouncer_stats_demo_pgbouncer_y2025w45 pgbouncer_stats_1  (cost=0.00..31.69 rows=246 width=32) (actual time=0.330..1.415 rows=251.00 loops=1)
               Filter: (dbname = 'demo_pgbouncer'::text)
               Buffers: shared hit=1 read=27 dirtied=1
               I/O Timings: shared read=0.831
         ->  Bitmap Heap Scan on pgbouncer_stats_demo_pgbouncer_y2025w46 pgbouncer_stats_2  (cost=4.17..11.29 rows=3 width=32) (actual time=0.010..0.010 rows=0.00 loops=1)
               Recheck Cond: (dbname = 'demo_pgbouncer'::text)
               Buffers: shared hit=2
               ->  Bitmap Index Scan on pgbouncer_stats_demo_pgbouncer_y2025w46_dbname_time_idx  (cost=0.00..4.17 rows=3 width=0) (actual time=0.005..0.006 rows=0.00 loops=1)
                     Index Cond: (dbname = 'demo_pgbouncer'::text)
                     Index Searches: 1
                     Buffers: shared hit=2
         ->  Bitmap Heap Scan on pgbouncer_stats_demo_pgbouncer_y2025w47 pgbouncer_stats_3  (cost=4.17..11.29 rows=3 width=32) (actual time=0.004..0.005 rows=0.00 loops=1)
               Recheck Cond: (dbname = 'demo_pgbouncer'::text)
               Buffers: shared hit=2
               ->  Bitmap Index Scan on pgbouncer_stats_demo_pgbouncer_y2025w47_dbname_time_idx  (cost=0.00..4.17 rows=3 width=0) (actual time=0.002..0.002 rows=0.00 loops=1)
                     Index Cond: (dbname = 'demo_pgbouncer'::text)
                     Index Searches: 1
                     Buffers: shared hit=2
         ->  Bitmap Heap Scan on pgbouncer_stats_demo_pgbouncer_y2025w48 pgbouncer_stats_4  (cost=4.17..11.29 rows=3 width=32) (actual time=0.011..0.012 rows=0.00 loops=1)
               Recheck Cond: (dbname = 'demo_pgbouncer'::text)
               Buffers: shared hit=2
               ->  Bitmap Index Scan on pgbouncer_stats_demo_pgbouncer_y2025w48_dbname_time_idx  (cost=0.00..4.17 rows=3 width=0) (actual time=0.004..0.005 rows=0.00 loops=1)
                     Index Cond: (dbname = 'demo_pgbouncer'::text)
                     Index Searches: 1
                     Buffers: shared hit=2
 Planning:
   Buffers: shared hit=3
 Planning Time: 0.311 ms
 Execution Time: 1.719 ms
(37 rows)

Index is used for panels but not for variable refresh. As you can see we need to check all partitions.

[slardiere]

Ok for selecting multiples values, but you need to change the SQL query with the database field in the window definition.

For the stat panels, perhaps use the sum of the values for the selected databases, instead of the repetitive panel, which I do not find very aesthetically pleasing. I can fix it if you agree.

[slardiere]

Indeed, adding time filter in the database selector use index : 

explain (analyse) select distinct (data->>'database') as database from pgbouncer_stats where dbname = 'pgb_spokane'   and time BETWEEN '2025-11-05T09:52:10.401Z' AND '2025-11-05T15:52:10.401Z' ;
                                                                                                           QUERY PLAN                                                                                              
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 HashAggregate  (cost=2552.06..2554.56 rows=200 width=32) (actual time=24.718..24.721 rows=5.00 loops=1)
   Group Key: (pgbouncer_stats.data ->> 'database'::text)
   Batches: 1  Memory Usage: 32kB
   Buffers: shared hit=1223
   ->  Index Scan using pgbouncer_stats_pgb_spokane_y2025w45_dbname_time_idx on pgbouncer_stats_pgb_spokane_y2025w45 pgbouncer_stats  (cost=0.29..2539.14 rows=5168 width=32) (actual time=0.121..19.925 rows=10795
         Index Cond: ((dbname = 'pgb_spokane'::text) AND ("time" >= '2025-11-05 10:52:10.401+01'::timestamp with time zone) AND ("time" <= '2025-11-05 16:52:10.401+01'::timestamp with time zone))
         Index Searches: 1
         Buffers: shared hit=1223
 Planning Time: 0.371 ms
 Execution Time: 24.763 ms

[0xgouda]

Update dynamically on time range change.

[pashagolub]

@slardiere would you please check the stats panels if they work for you? I'm fine with whatever option.

[pashagolub]

Good job @0xgouda! Thanks!

[slardiere]

@slardiere would you please check the stats panels if they work for you? I'm fine with whatever option.

@pashagolub yes, stats panels works for me, it's ok