Il nostro punto della strada

User Tools

Site Tools

Trace:
sysadmin:postgresql

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
Last revisionBoth sides next revision
sysadmin:postgresql [2014/11/05 09:40] – tabelle vuote ferodasysadmin:postgresql [2015/02/12 09:01] – Indice su espressione feroda
Line 1: Line 1:
-====== Basi di configurazione ======+====== PostgreSQL - Basi di configurazione ======
  
   - eseguire ''/usr/local/sbin/postgresql_shmall_shmmax.sh >> /etc/sysctl.conf''   - eseguire ''/usr/local/sbin/postgresql_shmall_shmmax.sh >> /etc/sysctl.conf''
Line 10: Line 10:
   * Performance Tuning: http://wiki.postgresql.org/wiki/Tuning_Your_PostgreSQL_Server   * Performance Tuning: http://wiki.postgresql.org/wiki/Tuning_Your_PostgreSQL_Server
   * Ottimo articolo segnalato da AndreaF http://www.revsys.com/writings/postgresql-performance.html   * Ottimo articolo segnalato da AndreaF http://www.revsys.com/writings/postgresql-performance.html
 +  * Index Maintenance https://wiki.postgresql.org/wiki/Index_Maintenance
  
  
 ===== Query utili su indici e loro utilizzo  ===== ===== Query utili su indici e loro utilizzo  =====
 +
 +Indici presenti
 +
 +    [0] SELECT * FROM pg_indexes;
  
 ==== Ricerca indici doppi ==== ==== Ricerca indici doppi ====
Line 32: Line 37:
     SELECT relid,schemaname,relname,idx_scan,idx_tup_read,idx_tup_fetch FROM pg_stat_all_indexes ORDER BY idx_scan desc;     SELECT relid,schemaname,relname,idx_scan,idx_tup_read,idx_tup_fetch FROM pg_stat_all_indexes ORDER BY idx_scan desc;
  
-Indici presenti 
- 
-    SELECT * FROM pg_indexes; 
-     
 Filtra solo quelli utilizzati almeno una volta Filtra solo quelli utilizzati almeno una volta
          
Line 64: Line 65:
 ===== Esecuzione query al ... ===== ===== Esecuzione query al ... =====
  
 +    * [0] = ... indici totali
     * [1] = ... indici doppi     * [1] = ... indici doppi
     * [2] = ... indici non usati     * [2] = ... indici non usati
Line 69: Line 71:
     * [4] = ... indici usati ma non efficaci     * [4] = ... indici usati ma non efficaci
     * [5] = ... tabelle vuote     * [5] = ... tabelle vuote
 +
 +====== CREAZIONE E USO INDICI =======
 +
 +  postgres# \d+ acs_service.connection_session;
 +     Column   |           Type           |                                  Modifiers                                  | Storage |                                                   Description                                                    
 +  ------------+--------------------------+-----------------------------------------------------------------------------+---------+------------------------------------------------------------------------------------------------------------------
 +   id         | integer                  | not null default nextval('acs_service.connection_session_id_seq'::regclass) | plain   | 
 +   start_time | timestamp with time zone | not null                                                                    | plain   | Moment in which the VPN session started. Obtained from $time_unix OpenVPN's envvar.
 +   duration   | interval                 |                                                                             | plain   | Duration of the connection, value is NULL until connection drops. Obtained from $time_duration OpenVPN's envvar.
 +   id_chamber | integer                  | not null                                                                    | plain   | Id of the Chamber related to the connection.
 +  Indexes:
 +    "pk_id" PRIMARY KEY, btree (id)
 +    "connection_session_duration_idx" btree (duration)
 +    "connection_session_id_chamber_idx" btree (id_chamber)
 +  Foreign-key constraints:
 +    "fk_chamber" FOREIGN KEY (id_chamber) REFERENCES acs_service.chamber(id)
 +  Has OIDs: no
 +
 +  postgres# REINDEX TABLE "acs_service"."connection_session";
 +  REINDEX
 +  postgres# EXPLAIN ANALYZE SELECT bcs.id, bcs.duration FROM acs_service.connection_session bcs WHERE bcs.duration IS NULL;
 +                                                                QUERY PLAN                                                                
 +  ------------------------------------------------------------------------------------------------------------------------------------------
 +  Bitmap Heap Scan on connection_session bcs  (cost=4.82..234.49 rows=70 width=20) (actual time=0.117..0.180 rows=67 loops=1)
 +   Recheck Cond: (duration IS NULL)
 +   ->  Bitmap Index Scan on connection_session_duration_idx  (cost=0.00..4.80 rows=70 width=0) (actual time=0.106..0.106 rows=67 loops=1)
 +         Index Cond: (duration IS NULL)
 +  Total runtime: 0.208 ms (5 rows)
 +
 +  postgres=# EXPLAIN ANALYZE SELECT bcs.id, bcs.duration FROM acs_service.connection_session bcs WHERE bcs.duration IS NOT NULL;
 +                                                          QUERY PLAN                                                           
 +  -------------------------------------------------------------------------------------------------------------------------------
 +   Seq Scan on connection_session bcs  (cost=0.00..2829.78 rows=149208 width=20) (actual time=0.008..36.858 rows=149211 loops=1)
 +   Filter: (duration IS NOT NULL)
 +  Total runtime: 46.523 ms (3 rows)
 +  
 +===== Se creo indice sull'espressione "duration IS NOT NULL" cosa succede? =====
 +
 +Niente .... :(
 +
 +    postgres=# CREATE INDEX ON "acs_service".connection_session ((duration IS NOT NULL));
 +    CREATE INDEX
 +    postgres=# REINDEX TABLE "acs_service"."connection_session";
 +    REINDEX
 +    postgres=# EXPLAIN ANALYZE SELECT bcs.id, bcs.duration FROM acs_service.connection_session bcs WHERE bcs.duration IS NOT NULL;
 +                                                              QUERY PLAN                                                       
 +    -------------------------------------------------------------------------------------------------------------------------------
 +     Seq Scan on connection_session bcs  (cost=0.00..2829.90 rows=149220 width=20) (actual time=0.011..38.323 rows=149226 loops=1)
 +       Filter: (duration IS NOT NULL)
 +     Total runtime: 48.491 ms
 +    (3 rows)
 +
 +    postgres=# EXPLAIN ANALYZE SELECT bcs.id, bcs.duration FROM acs_service.connection_session bcs WHERE bcs.duration IS NULL;         
 +                                                                    QUERY PLAN                                                            
 +    ------------------------------------------------------------------------------------------------------------------------------------------
 +     Bitmap Heap Scan on connection_session bcs  (cost=4.82..234.49 rows=70 width=20) (actual time=0.061..0.119 rows=65 loops=1)
 +       Recheck Cond: (duration IS NULL)
 +       ->  Bitmap Index Scan on connection_session_duration_idx  (cost=0.00..4.80 rows=70 width=0) (actual time=0.050..0.050 rows=66 loops=1)
 +             Index Cond: (duration IS NULL)
 +     Total runtime: 0.148 ms
 +    (5 rows)
 +
 +
 +====== Occupazione disco ======
 +
 +===== Database =====
 +
 +Lo restituisce pure in formato carino :)
 +
 +# select (pg_database_size(current_database())/(1024*1024))::text || ' MB' as db_size;
 +
 +===== Tabelle =====
 +
 +Restituisce un elenco ordinato in modo discendente di:
 +
 +  * schema (aka "directory")
 +  * tabella
 +  * occupazione su disco
 +  * occupazione totale compresi indici
 +  * numero stimato di righe
 +
 +    SELECT table_schema, table_name, 
 +      (size/1024)::varchar || ' kB' as size,
 +      (total_size/1024)::varchar || ' kB' as total_size,
 +      (CASE WHEN pg_class.reltuples > 0 THEN
 +          basic_infos.size/(8192*relpages*reltuples)
 +          ELSE 0
 +          END
 +      )::bigint AS estimated_row_count
 +      FROM (
 +        SELECT table_schema, table_name, 
 +            pg_relation_size('"' || table_schema || '"."' || table_name || '"') as size,
 +            pg_total_relation_size('"' || table_schema || '"."' || table_name || '"') as total_size
 +            FROM information_schema.tables 
 +            WHERE table_schema NOT IN ('information_schema', 'pg_catalog') 
 +        ORDER BY size DESC, total_size DESC
 +    ) AS basic_infos 
 +        JOIN pg_class
 +        ON ('"' || table_schema || '"."' || table_name || '"')::regclass = pg_class.oid
 +
 +Questa query funziona su PostgreSQL 8.4. E credo valga anche per PostgreSQL >= 9.0. Potrebbe essere interessante pubblicarla da qualche parte.
 +
 +
 +====== Informazioni varie... ======
  
 ===== Query comode x debug ===== ===== Query comode x debug =====