Adding --exclude support for last_vacuum and last_analyze

[SkyMarshall939]

Hello,

Here is a proposal for improvements to the last_vacuum & last_analyze services. The purpose is to add support for the --exclude parameter for these services in order to be able to exclude X schemas and/or X relations.

Changes

=> Change NaN to -1 to allow counters to be accumulated in Perl code.
=> Add current_database(), schemaname and relname in the last_[analyze|vacuum] query.
=> Relations with n_live_tup = 0 are no longer taken.

How

Classic use of the service concerned by adding the --exclude parameter which takes REGEXP .

# Exclusion of temp and tmp schemas 
check_pgactivity -s last_vacuum --username dbuser --host myhost --dbname mydb --warning 30d --critical 45d --port 5433 --exclude '.*.temp\|tmp..*'

I remain available for any exchange,

Best regards,

[robinportigliatti]

Relations with n_live_tup = 0 are no longer taken.

It should be taken into considerations. What you don't want is a partition_table (the mother) = 0

[blogh]

I think skipping table with zero tuple is a regression. But the option would be nice.

As for the partition thing mentionned by @robinportigliatti, 'Declarative Partition"
cannot be vacuumed, but are still catched by this service. But there is an issue for that : #365. It think, it should have it's own PR.

[[local]:5434 (16.8)] postgres@postgres=# CREATE TABLE ttt (i int) PARTITION BY LIST (i);                                              
CREATE TABLE                                                                                                                           
[[local]:5434 (16.8)] postgres@postgres=# CREATE TABLE ttt_1 PARTITION OF ttt FOR VALUES IN (1,2,3);                                   
CREATE TABLE                                                                                                                           
[[local]:5434 (16.8)] postgres@postgres=# SELECT relname, last_vacuum, vacuum_count FROM pg_stat_user_tables WHERE relname LIKE 'ttt%';
 relname | last_vacuum | vacuum_count                                                                                                  
---------+-------------+--------------                                                                                                 
 ttt     | ø           |            0                                                                                                  
 ttt_1   | ø           |            0                                                                                                  
(2 rows)                                                                                                                               
                                                                                                                                       
[[local]:5434 (16.8)] postgres@postgres=# VACUUM ttt;                                                                                  
VACUUM                                                                                                                                 
[[local]:5434 (16.8)] postgres@postgres=# SELECT relname, last_vacuum, vacuum_count FROM pg_stat_user_tables WHERE relname LIKE 'ttt%';
 relname |          last_vacuum          | vacuum_count                                                                                
---------+-------------------------------+--------------                                                                               
 ttt     | ø                             |            0                                                                                
 ttt_1   | 2025-05-19 08:34:50.655983+00 |            1                                                                                
(2 rows)                                                                                                                               

[SkyMarshall939]

To keep things simple and clean, I remove the n_live_tup condition and focus the patch on pattern management.

Thanks for your return,

Regards,

[blogh]

I don't think this is the proper way to specify this.
@ioguix ?

[blogh]

When maintenance_max=Infinity we display Infinitys.

[gleu]

This needs to be fixed. Can you fix this @SkyMarshall939 ?

[blogh]

@ioguix @Krysztophe That's about all I can do review-wise in perl.

[gleu]

It needs a rebase.

[blogh]

rebased

[blogh]

@SkyMarshall939 If you want to skim through the patch again ... that would be helpful (I don't think I broke anything else but me and perl ...).

which kinda hints at the need for tests for this patch

[blogh]

just found another few dumb mistakes. will post a fix soon.

[blogh]

Ok ... fixed ... for good .. hopefully.

I feel dumb, I should have tested.

Also, I feel dirty: I discovered autovivification (accessing an empty nested struct creates it): this is pure madness.