Re: Vacuum statistics - Mailing list pgsql-hackers

Hi!

On 16.10.2024 14:01, Alena Rybakina wrote:

Thank you for rebasing.

I have noticed that when I create a table or an index on this table, there is no information about the table or index in pg_stat_vacuum_tables and pg_stat_vacuum_indexes until we perform a VACUUM.

Example:

CREATE TABLE t (i INT, j INT);
INSERT INTO t SELECT i/10, i/100 FROM  GENERATE_SERIES(1,1000000) i;
SELECT * FROM pg_stat_vacuum_tables WHERE relname = 't';
....
(0 rows)
CREATE INDEX ON t (i);
SELECT * FROM pg_stat_vacuum_indexes WHERE relname = 't_i_idx';
...
(0 rows)

I can see the entries after running VACUUM or executing autovacuum. or when autovacuum is executed. I would suggest adding a line about the relation even if it has not yet been processed by vacuum. Interestingly, this issue does not occur with pg_stat_vacuum_database:

CREATE DATABASE example_db;
SELECT * FROM pg_stat_vacuum_database WHERE dbname = 'example_db';
dboid |       dbname | ...
 ...      | example_db | ...
(1 row)

BTW, I recommend renaming the view pg_stat_vacuum_database to pg_stat_vacuum_databaseS  for consistency with pg_stat_vacuum_tables and pg_stat_vacuum_indexes

Thanks for the review. I'm investigating this. I agree with the renaming, I will do it in the next version of the patch.

I fixed it. I added the left outer join to the vacuum views and for converting the coalesce function from NULL to null values.

I also fixed the code in getting database statistics - we can get it through the existing pgstat_fetch_stat_dbentry function and fixed couple of comments. 

I attached a diff file, as well as new versions of patches.

-- 
Regards,
Alena Rybakina
Postgres Professional

Thank you for fixing it.

1) I have found some typos in the test output files (out-files) when running 'make check' and 'make check-world'. These typos might cause minor discrepancies in test results. You may already be aware of them, but I wanted to bring them to your attention in case they haven't been noticed. I believe these can be fixed quickly.

2) Additionally, I observed that when we create a table and insert some rows, executing the VACUUM FULL command does not update the information in the 'pg_stat_get_vacuum_tables' However, running the VACUUM command does update this information as expected. This seems inconsistent, and it might be a bug.

Example:
CREATE TABLE t (i INT, j INT) WITH (autovacuum_enabled = false);
INSERT INTO t SELECT i/10, i/100 FROM  GENERATE_SERIES(1,1000000) i;
SELECT * FROM pg_stat_get_vacuum_tables WHERE relname = 't';
schema | relname |    relid | total_blks_read | .........
-----------+------------+---------+----------------------+---------
   public | t            | 21416 |                       0 | ......
(1 row)

VACUUM FULL;
SELECT * FROM pg_stat_get_vacuum_tables WHERE relname = 't';
schema | relname |    relid | total_blks_read | .........
-----------+------------+---------+----------------------+---------
   public | t            | 21416 |                       0 | ......
(1 row)

VACUUM;
SELECT * FROM pg_stat_get_vacuum_tables WHERE relname = 't';
schema | relname |    relid | total_blks_read | .........
-----------+------------+---------+----------------------+---------
   public | t            | 21416 |                 4425 | ......
(1 row)

Regards,
Ilia Evdokimov,
Tantor Labs LLC.