Continuação do artigo "«.
O artigo abordará e mostrará, com consultas específicas e exemplos, qual informação útil pode ser obtida através do histórico da visão pg_stat_activity.
Aviso.
Devido à novidade do tema e à conclusão não finalizada do período de testes, o artigo pode conter erros. Críticas e comentários são sempre bem-vindos e esperados.
Dados de entrada
História da visão pg_stat_statements
pg_stat_history
CREATE TABLE pg_stat_history (
id SERIAL,
snapshot_timestamp timestamp sem fuso horário,
database_id integer,
dbid oid,
userid oid,
queryid bigint,
query text,
calls bigint,
total_time double precision,
min_time double precision,
max_time double precision,
mean_time double precision,
stddev_time double precision,
rows bigint,
shared_blks_hit bigint,
shared_blks_read bigint,
shared_blks_dirtied bigint,
shared_blks_written bigint,
local_blks_hit bigint,
local_blks_read bigint,
local_blks_dirtied bigint,
local_blks_written bigint,
temp_blks_read bigint,
temp_blks_written bigint,
blk_read_time double precision,
blk_write_time double precision,
baseline_id integer );A tabela é preenchida a cada hora, utilizando dblink para o banco de dados de destino. A coluna mais interessante e útil na tabela, é claro, queryid.
História da visão pg_stat_activity
archive_pg_stat_activity
CREATE TABLE archive_pg_stat_activity
(
timepoint timestamp sem fuso horário,
datid oid,
datname name,
pid integer,
usesysid oid,
usename name,
application_name text,
client_addr inet,
client_hostname text,
client_port integer,
backend_start timestamp sem fuso horário,
xact_start timestamp sem fuso horário,
query_start timestamp sem fuso horário,
state_change timestamp sem fuso horário,
wait_event_type text,
wait_event text,
state text,
backend_xid xid,
backend_xmin xid,
query text,
backend_type text,
queryid bigint
);A tabela é uma tabela particionada por horas da história_pg_stat_activity (Mais detalhes aqui — e aqui —
Saída
TEMPO DE CPU AGRUPADO (SISTEMA + CLIENTES)
Consulta
WITH
t AS
(
SELECT
date_trunc('second', timepoint)
FROM activity_hist.archive_pg_stat_activity aa
WHERE timepoint BETWEEN pg_stat_history_begin+(current_hour_diff * interval '1 hour') AND pg_stat_history_end+(current_hour_diff * interval '1 hour') AND
( aa.wait_event_type IS NULL ) AND
aa.state = 'active'
)
SELECT count(*)
INTO cpu_total
FROM t ;Exemplo
TEMPO DE CPU AGRUPADO (SISTEMA + CLIENTES) : 28:37:46TEMPOS DE ESPERA AGRUPADOS
Consulta
WITH
t AS
(
SELECT
date_trunc('second', timepoint)
FROM activity_hist.archive_pg_stat_activity aa
WHERE timepoint BETWEEN pg_stat_history_begin+(current_hour_diff * interval '1 hour') AND pg_stat_history_end+(current_hour_diff * interval '1 hour') AND
( aa.wait_event_type IS NOT NULL ) AND
aa.state = 'active'
)
SELECT count(*)
INTO cpu_total
FROM t ;Exemplo
TEMPOS DE ESPERA AGRUPADOS : 30:12:49Total de valores do pg_stat_statements
Consulta
--TOTAL pg_stat
SELECT
SUM(calls) AS calls, SUM(total_time) AS total_time, SUM(rows) AS rows ,
SUM(shared_blks_hit) AS shared_blks_hit,SUM(shared_blks_read) AS shared_blks_read ,
SUM(shared_blks_dirtied) AS shared_blks_dirtied,SUM(shared_blks_written) AS shared_blks_written ,
SUM(local_blks_hit) AS local_blks_hit , SUM(local_blks_read) AS local_blks_read ,
SUM(local_blks_dirtied) AS local_blks_dirtied , SUM(local_blks_written) AS local_blks_written,
SUM(temp_blks_read) AS temp_blks_read, SUM(temp_blks_written) temp_blks_written ,
SUM(blk_read_time) AS blk_read_time , SUM(blk_write_time) AS blk_write_time
INTO
pg_total_stat_history_rec
FROM
pg_stat_history
WHERE
snapshot_timestamp BETWEEN pg_stat_history_begin AND pg_stat_history_end AND
queryid IS NULL;SQL DBTIME — tempo total de execução de consultas
Consulta
dbtime_total = intervalo '1 milissegundo' * pg_total_stat_history_rec.total_time ;Exemplo
SQL DBTIME : 136:49:36SQL CPU TIME — tempo de CPU gasto na execução de consultas
Consulta
COM COM
t COMO
(
SELECT
date_trunc('second', timepoint)
FROM activity_hist.archive_pg_stat_activity aa
WHERE timepoint BETWEEN pg_stat_history_begin+(current_hour_diff * intervalo '1 hour') AND pg_stat_history_end+(current_hour_diff * intervalo '1 hour') AND
( aa.wait_event_type IS NULL ) AND
backend_type = 'client backend' AND
aa.state = 'active'
)
SELECT count(*)
INTO cpu_total
FROM t ;Exemplo
SQL CPU TIME : 27:40:15SQL WAITINGS TIME — tempo total de espera para consultas
Consulta
COM COM
t COMO
(
SELECT
date_trunc('second', timepoint)
FROM activity_hist.archive_pg_stat_activity aa
WHERE timepoint BETWEEN pg_stat_history_begin+(current_hour_diff * intervalo '1 hour') AND pg_stat_history_end+(current_hour_diff * intervalo '1 hour') AND
( aa.wait_event_type IS NOT NULL ) AND
aa.state = 'active' AND
backend_type = 'client backend'
)
SELECT count(*)
INTO waiting_total
FROM t ;Exemplo
SQL WAITINGS TIME : 30:04:09As consultas a seguir são triviais e, para economizar espaço, os detalhes da implementação são omitidos:
Exemplo
| SQL IOTIME : 19:44:50
| SQL READ TIME : 19:44:32
| SQL WRITE TIME : 00:00:17
|
| SQL CALLS : 12188248
-------------------------------------------------------------
| SQL SHARED BLOCKS READS : 7997039120
| SQL SHARED BLOCKS HITS : 8868286092
| SQL SHARED BLOCKS HITS/READS % : 110.89
| SQL SHARED BLOCKS DIRTIED : 419945
| SQL SHARED BLOCKS WRITTEN : 19857
|
| SQL TEMPORARY BLOCKS READS : 7836169
| SQL TEMPORARY BLOCKS WRITTEN : 10683938
Vamos para a seção mais interessante
ESTATÍSTICAS DE ESPERA
TOP 10 WAITINGS BY TOTAL WAIT TIME FOR CLIENTS PROCESSES
Consulta
SELECT
wait_event_type , wait_event ,
get_system_waiting_duration( wait_event_type , wait_event ,pg_stat_history_begin+(current_hour_diff * intervalo '1 hour') ,pg_stat_history_end+(current_hour_diff * intervalo '1 hour') ) as duration
FROM
activity_hist.archive_pg_stat_activity aa
WHERE
timepoint BETWEEN pg_stat_history_begin+(current_hour_diff * intervalo '1 hour') AND pg_stat_history_end+(current_hour_diff * intervalo '1 hour') AND backend_type != 'client backend' AND wait_event_type IS NOT NULL
GROUP BY
wait_event_type, wait_event
ORDER BY 3 DESC
LIMIT 10Exemplo
+------------------------------------------------------------------------------------ | TOP 10 AGUARDANDO PELO TEMPO TOTAL DE ESPERA DOS PROCESSOS DO SISTEMA +-----+------------------------------+--------------------+-------------------- | #| tipo_evento_espera| evento_espera| duração +-----+------------------------------+--------------------+-------------------- | 1| Atividade| LogicalLauncherMain| 10:43:28 | 2| Atividade| AutoVacuumMain| 10:42:49 | 3| Atividade| WalWriterMain| 10:28:53 | 4| Atividade| CheckpointerMain| 10:23:50 | 5| Atividade| BgWriterMain| 09:11:59 | 6| Atividade| BgWriterHibernate| 01:37:46 | 7| IO| BufFileWrite| 00:02:35 | 8| LWLock| buffer_mapping| 00:01:54 | 9| IO| DataFileRead| 00:01:23 | 10| IO| WALWrite| 00:00:59 +-----+------------------------------+--------------------+--------------------
TOP 10 WAITINGS BY TOTAL WAIT TIME FOR CLIENTS PROCESSES
Consulta
SELECIONE
tipo_evento_espera , evento_espera ,
get_clients_waiting_duration( tipo_evento_espera , evento_espera , pg_stat_history_begin+(current_hour_diff * interval '1 hour') , pg_stat_history_end+(current_hour_diff * interval '1 hour') ) as duração
DE
activity_hist.archive_pg_stat_activity aa
ONDE
timepoint ENTRE pg_stat_history_begin+(current_hour_diff * interval '1 hour') E pg_stat_history_end+(current_hour_diff * interval '1 hour') E tipo_backend = 'client backend' E tipo_evento_espera NÃO É NULO
AGRUPAR POR tipo_evento_espera, evento_espera
ORDER BY 3 DESC
LIMIT 10Exemplo
+-----+------------------------------+--------------------+--------------------+---------- | #| tipo_evento_espera| evento_espera| duração| % dbtime +-----+------------------------------+--------------------+--------------------+---------- | 1| Bloqueio| transactionid| 08:16:47| 6.05 | 2| IO| DataFileRead| 06:13:41| 4.55 | 3| Timeout| PgSleep| 02:53:21| 2.11 | 4| LWLock| buffer_mapping| 00:40:42| 0.5 | 5| LWLock| buffer_io| 00:17:17| 0.21 | 6| IO| BufFileWrite| 00:01:34| 0.02 | 7| Bloqueio| tuple| 00:01:32| 0.02 | 8| Cliente| ClientRead| 00:01:19| 0.02 | 9| IO| BufFileRead| 00:00:37| 0.01 | 10| LWLock| buffer_content| 00:00:08| 0 +-----+------------------------------+--------------------+--------------------+----------
TIPOS DE ESPERA PELO TEMPO TOTAL DE ESPERA, PARA PROCESSOS DO SISTEMA
Consulta
SELECIONE
tipo_evento_espera ,
get_system_waiting_type_duration( tipo_evento_espera , pg_stat_history_begin+(current_hour_diff * interval '1 hora') , pg_stat_history_end+(current_hour_diff * interval '1 hora') ) as duração
DE
activity_hist.archive_pg_stat_activity aa
ONDE
timepoint ENTRE pg_stat_history_begin+(current_hour_diff * interval '1 hora') E pg_stat_history_end+(current_hour_diff * interval '1 hora') E tipo_backend != 'client backend' E tipo_evento_espera NÃO É NULO
AGRUPAR POR tipo_evento_espera
ORDER BY 2 DESCExemplo
+-----+------------------------------+-------------------- | #| tipo_de_evento_de_espera| duração +-----+------------------------------+-------------------- | 1| Atividade| 53:08:45 | 2| IO| 00:06:24 | 3| LWLock| 00:03:02 +-----+------------------------------+--------------------
TIPOS DE ESPERA POR TEMPO TOTAL DE ESPERA, PARA PROCESSES DE CLIENTE
Consulta
SELECIONE
tipo_de_evento_de_espera ,
get_clients_waiting_type_duration( tipo_de_evento_de_espera , pg_stat_history_begin+(current_hour_diff * interval '1 hour') , pg_stat_history_end+(current_hour_diff * interval '1 hour') ) como duração
DE
activity_hist.archive_pg_stat_activity aa
ONDE
ponto_de_tempo ENTRE pg_stat_history_begin+(current_hour_diff * interval '1 hour') E pg_stat_history_end+(current_hour_diff * interval '1 hour') E backend_type = 'client backend' E tipo_de_evento_de_espera NÃO É NULO
AGRUPAR POR tipo_de_evento_de_espera
ORDER BY 2 DESCExemplo
+-----+------------------------------+--------------------+-------------------- | #| tipo_de_evento_de_espera| duração| % dbtime +-----+------------------------------+--------------------+-------------------- | 1| Lock| 08:18:19| 6.07 | 2| IO| 06:16:01| 4.58 | 3| Timeout| 02:53:21| 2.11 | 4| LWLock| 00:58:12| 0.71 | 5| Cliente| 00:01:19| 0.02 | 6| IPC| 00:00:04| 0 +-----+------------------------------+--------------------+--------------------
Durações de espera, para processos do sistema e solicitações individuais.
ESPERAS PARA PROCESSOS DO SISTEMA
Consulta
SELECIONE
backend_type , datname , tipo_de_evento_de_espera , wait_event , get_backend_type_waiting_duration( backend_type , tipo_de_evento_de_espera , wait_event , pg_stat_history_begin+(current_hour_diff * interval '1 hour') , pg_stat_history_end+(current_hour_diff * interval '1 hour') ) como duração
DE
activity_hist.archive_pg_stat_activity aa
ONDE
ponto_de_tempo ENTRE pg_stat_history_begin+(current_hour_diff * interval '1 hour') E pg_stat_history_end+(current_hour_diff * interval '1 hour') E backend_type != 'client backend' E tipo_de_evento_de_espera NÃO É NULO
AGRUPAR POR backend_type , datname , tipo_de_evento_de_espera , wait_event
ORDER BY 5 DESCExemplo
+-----+-----------------------------+----------+--------------------+----------------------+-------------------- | #| backend_type| dbname| wait_event_type| wait_event| duration +-----+-----------------------------+----------+--------------------+----------------------+-------------------- | 1| lançador de replicação lógica| | Atividade| LogicalLauncherMain| 10:43:28 | 2| lançador de autovacuum| | Atividade| AutoVacuumMain| 10:42:49 | 3| walwriter| | Atividade| WalWriterMain| 10:28:53 | 4| checkpointer| | Atividade| CheckpointerMain| 10:23:50 | 5| escritor em segundo plano| | Atividade| BgWriterMain| 09:11:59 | 6| escritor em segundo plano| | Atividade| BgWriterHibernate| 01:37:46 | 7| trabalhador paralelo| tdb1| IO| BufFileWrite| 00:02:35 | 8| trabalhador paralelo| tdb1| LWLock| buffer_mapping| 00:01:41 | 9| trabalhador paralelo| tdb1| IO| DataFileRead| 00:01:22 | 10| trabalhador paralelo| tdb1| IO| BufFileRead| 00:00:59 | 11| walwriter| | IO| WALWrite| 00:00:57 | 12| trabalhador paralelo| tdb1| LWLock| buffer_io| 00:00:47 | 13| trabalhador de autovacuum| tdb1| LWLock| buffer_mapping| 00:00:13 | 14| escritor em segundo plano| | IO| DataFileWrite| 00:00:12 | 15| checkpointer| | IO| DataFileWrite| 00:00:11 | 16| walwriter| | LWLock| WALWriteLock| 00:00:09 | 17| checkpointer| | LWLock| WALWriteLock| 00:00:06 | 18| escritor em segundo plano| | LWLock| WALWriteLock| 00:00:06 | 19| walwriter| | IO| WALInitWrite| 00:00:02 | 20| trabalhador de autovacuum| tdb1| LWLock| WALWriteLock| 00:00:02 | 21| walwriter| | IO| WALInitSync| 00:00:02 | 22| trabalhador de autovacuum| tdb1| IO| DataFileRead| 00:00:01 | 23| checkpointer| | IO| ControlFileSyncUpdate| 00:00:01 | 24| escritor em segundo plano| | IO| WALWrite| 00:00:01 | 25| escritor em segundo plano| | IO| DataFileFlush| 00:00:01 | 26| checkpointer| | IO| SLRUFlushSync| 00:00:01 | 27| trabalhador de autovacuum| tdb1| IO| WALWrite| 00:00:01 | 28| checkpointer| | IO| DataFileSync| 00:00:01 +-----+-----------------------------+----------+--------------------+----------------------+--------------------
WAITINGS FOR SQL — espera para consultas específicas por queryid
Consulta
SELECIONE
queryid, datname, wait_event_type, wait_event, get_query_waiting_duration(queryid, wait_event_type, wait_event, pg_stat_history_begin+(current_hour_diff * interval '1 hour'), pg_stat_history_end+(current_hour_diff * interval '1 hour')) AS duration
FROM
activity_hist.archive_pg_stat_activity aa
WHERE
timepoint ENTRE pg_stat_history_begin+(current_hour_diff * interval '1 hora') E pg_stat_history_end+(current_hour_diff * interval '1 hora') E backend_type = 'client backend' E wait_event_type IS NOT NULL E queryid IS NOT NULL
AGRUPAR POR queryid, datname, wait_event_type, wait_event
ORDER BY 1, 5 DESC Exemplo
+-----+-------------------------+----------+--------------------+--------------------+--------------------+-------------------- | #| queryid| dbname| wait_event_type| wait_event| waitings| total | | | | | | duration| duration +-----+-------------------------+----------+--------------------+--------------------+--------------------+-------------------- | 1| -8247416849404883188| tdb1| Cliente| ClientRead| 00:00:02| | 2| -6572922443698419129| tdb1| Cliente| ClientRead| 00:00:05| | 3| -6572922443698419129| tdb1| IO| DataFileRead| 00:00:01| | 4| -5917408132400665328| tdb1| Cliente| ClientRead| 00:00:04| | 5| -4091009262735781873| tdb1| Cliente| ClientRead| 00:00:03| | 6| -1473395109729441239| tdb1| Cliente| ClientRead| 00:00:01| | 7| 28942442626229688| tdb1| IO| BufFileWrite| 00:01:34| 00:46:06 | 8| 28942442626229688| tdb1| LWLock| buffer_mapping| 00:01:05| 00:46:06 | 9| 28942442626229688| tdb1| IO| DataFileRead| 00:00:44| 00:46:06 | 10| 28942442626229688| tdb1| IO| BufFileRead| 00:00:37| 00:46:06 | 11| 28942442626229688| tdb1| LWLock| buffer_io| 00:00:35| 00:46:06 | 12| 28942442626229688| tdb1| Cliente| ClientRead| 00:00:05| 00:46:06 | 13| 28942442626229688| tdb1| IPC| MessageQueueReceive| 00:00:03| 00:46:06 | 14| 28942442626229688| tdb1| IPC| BgWorkerShutdown| 00:00:01| 00:46:06 | 15| 389015618226997618| tdb1| Lock| transactionid| 03:55:09| 04:14:15 | 16| 389015618226997618| tdb1| IO| DataFileRead| 03:23:09| 04:14:15 | 17| 389015618226997618| tdb1| LWLock| buffer_mapping| 00:12:09| 04:14:15 | 18| 389015618226997618| tdb1| LWLock| buffer_io| 00:10:18| 04:14:15 | 19| 389015618226997618| tdb1| Lock| tuple| 00:00:35| 04:14:15 | 20| 389015618226997618| tdb1| LWLock| WALWriteLock| 00:00:02| 04:14:15 | 21| 389015618226997618| tdb1| IO| DataFileWrite| 00:00:01| 04:14:15 | 22| 389015618226997618| tdb1| LWLock| SyncScanLock| 00:00:01| 04:14:15 | 23| 389015618226997618| tdb1| Cliente| ClientRead| 00:00:01| 04:14:15 | 24| 734234407411547467| tdb1| Cliente| ClientRead| 00:00:11| | 25| 734234407411547467| tdb1| LWLock| buffer_mapping| 00:00:05| | 26| 734234407411547467| tdb1| IO| DataFileRead| 00:00:02| | 27| 1237430309438971376| tdb1| LWLock| buffer_mapping| 00:02:18| 02:45:40 | 28| 1237430309438971376| tdb1| IO| DataFileRead| 00:00:27| 02:45:40 | 29| 1237430309438971376| tdb1| Cliente| ClientRead| 00:00:02| 02:45:40 | 30| 2404820632950544954| tdb1| Cliente| ClientRead| 00:00:01| | 31| 2515308626622579467| tdb1| Cliente| ClientRead| 00:00:02| | 32| 4710212362688288619| tdb1| LWLock| buffer_mapping| 00:03:08| 02:18:21 | 33| 4710212362688288619| tdb1| IO| DataFileRead| 00:00:22| 02:18:21 | 34| 4710212362688288619| tdb1| Cliente| ClientRead| 00:00:06| 02:18:21 | 35| 4710212362688288619| tdb1| LWLock| buffer_io| 00:00:02| 02:18:21 | 36| 9150846928388977274| tdb1| IO| DataFileRead| 00:01:19| | 37| 9150846928388977274| tdb1| LWLock| buffer_mapping| 00:00:34| | 38| 9150846928388977274| tdb1| Cliente| ClientRead| 00:00:10| | 39| 9150846928388977274| tdb1| LWLock| buffer_io| 00:00:01| +-----+-------------------------+----------+--------------------+--------------------+--------------------+--------------------
ESTATÍSTICAS SQL DO CLIENTE — PRINCIPAIS CONSULTAS
As consultas para obtenção continuam sendo triviais e, para economizar espaço, não são apresentadas.
Exemplos
+------------------------------------------------------------------------------------ | CLIENT SQL ordenado pelo Tempo Decorrido +--------------------+----------+----------+----------+----------+----------+-------------------- | tempo decorrido| chamadas| % dbtime| % CPU| % IO| dbname| queryid +--------------------+----------+----------+----------+----------+----------+-------------------- | 04:14:15| 19| 3.1| 10.83| 11.52| tdb1| 389015618226997618 | 02:45:40| 746| 2.02| 4.23| 0.08| tdb1| 1237430309438971376 | 02:18:21| 749| 1.69| 3.39| 0.1| tdb1| 4710212362688288619 | 00:46:06| 375| 0.56| 0.94| 0.41| tdb1| 28942442626229688 +--------------------+----------+----------+----------+----------+----------+-------------------- | CLIENT SQL ordenado pelo Tempo de CPU +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | tempo cpu| chamadas| % dbtime|tempo_total| % CPU| % IO| dbname| queryid +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | 02:59:49| 19| 3.1| 04:14:15| 10.83| 11.52| tdb1| 389015618226997618 | 01:10:12| 746| 2.02| 02:45:40| 4.23| 0.08| tdb1| 1237430309438971376 | 00:56:15| 749| 1.69| 02:18:21| 3.39| 0.1| tdb1| 4710212362688288619 | 00:15:35| 375| 0.56| 00:46:06| 0.94| 0.41| tdb1| 28942442626229688 +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | CLIENT SQL ordenado pelo Tempo de Espera de I/O do Usuário +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | tempo_io| chamadas| % dbtime|tempo_total| % CPU| % IO| dbname| queryid +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | 03:23:10| 19| 3.1| 04:14:15| 10.83| 11.52| tdb1| 389015618226997618 | 00:02:54| 375| 0.56| 00:46:06| 0.94| 0.41| tdb1| 28942442626229688 | 00:00:27| 746| 2.02| 02:45:40| 4.23| 0.08| tdb1| 1237430309438971376 | 00:00:22| 749| 1.69| 02:18:21| 3.39| 0.1| tdb1| 4710212362688288619 +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | CLIENT SQL ordenado pelas Leituras de Buffers Compartilhados +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | leituras de buffers| chamadas| % dbtime|tempo_total| % CPU| % IO| dbname| queryid +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | 1056388566| 19| 3.1| 04:14:15| 10.83| 11.52| tdb1| 389015618226997618 | 11709251| 375| 0.56| 00:46:06| 0.94| 0.41| tdb1| 28942442626229688 | 3439004| 746| 2.02| 02:45:40| 4.23| 0.08| tdb1| 1237430309438971376 | 3373330| 749| 1.69| 02:18:21| 3.39| 0.1| tdb1| 4710212362688288619 +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | CLIENT SQL ordenado pelo Tempo de Leituras de Disco +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | tempo leitura| chamadas| % dbtime|tempo_total| % CPU| % IO| dbname| queryid +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | 02:16:30| 19| 3.1| 04:14:15| 10.83| 11.52| tdb1| 389015618226997618 | 00:04:50| 375| 0.56| 00:46:06| 0.94| 0.41| tdb1| 28942442626229688 | 00:01:10| 749| 1.69| 02:18:21| 3.39| 0.1| tdb1| 4710212362688288619 | 00:00:57| 746| 2.02| 02:45:40| 4.23| 0.08| tdb1| 1237430309438971376 +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | CLIENT SQL ordenado pelo Número de Execuções +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | chamadas| linhas| % dbtime|tempo_total| % CPU| % IO| dbname| queryid +--------------------+----------+----------+----------+----------+----------+----------+-------------------- | 749| 749| 1.69| 02:18:21| 3.39| 0.1| tdb1| 4710212362688288619 | 746| 746| 2.02| 02:45:40| 4.23| 0.08| tdb1| 1237430309438971376 | 375| 0| 0.56| 00:46:06| 0.94| 0.41| tdb1| 28942442626229688 | 19| 19| 3.1| 04:14:15| 10.83| 11.52| tdb1| 389015618226997618 +--------------------+----------+----------+----------+----------+----------+----------+--------------------
Resultado
Usando as consultas apresentadas e os relatórios obtidos, é possível obter uma visão mais completa para análise e resolução de problemas de degradação de desempenho de consultas individuais e de todo o cluster, como um todo.
Desenvolvimento
Os planos de desenvolvimento atuais são os seguintes:
- Adicionar ao relatório um histórico de bloqueios. As consultas estão sendo testadas e serão apresentadas em breve.
- Utilizar a extensão TimescaleDB para armazenar o histórico de pg_stat_activity e pg_locks.
- Preparar uma solução em pacote no GitHub para implantação em massa em bancos de dados de produção.
Continua…
Fonte: habr.com
