Question

我有一个非常简单的数据库，我用它为Elasticsearch构建一些JSON。我是唯一连接到数据库的人，并且仅将其用于此特定任务。无论如何，在数据库上点击一些表格是非常慢的，这么慢，以至于构建我正在研究的Elasticsearch索引需要几个月的时间。我查看了有关Postgresql调优等的所有文章，而我正在改变的是解决问题。正如您在EXPLAIN ANALYZE部分中看到的那样，我正在进行的查询是一个非常简单的查询，没有连接或其他任何东西，而且它仍然非常慢。我只想弄清楚我能做些什么来加快速度，因为我认为没有理由这么慢。

以下是我能想到的所有相关信息，如果需要更多信息，我可以补充一下。

系统规格

[root@pgdb ~]# free -m
             total       used       free     shared    buffers     cached
Mem:         61444      39416      22027        458          8      38531
-/+ buffers/cache:        876      60568
Swap:            0          0          0


[root@pgdb ~]# cat /proc/cpuinfo
processor   : 0
vendor_id   : GenuineIntel
cpu family  : 6
model       : 62
model name  : Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz

所有表格

cotitsk=# \d+
                          List of relations
 Schema |       Name       | Type  |  Owner   |  Size   | Description
--------+------------------+-------+----------+---------+-------------
 public | company          | table | postgres | 1838 MB |
 public | company_industry | table | postgres | 1621 MB |
 public | company_title    | table | postgres | 3837 MB |
 public | industry         | table | postgres | 224 kB  |
 public | industry_skill   | table | postgres | 446 MB  |
 public | industry_title   | table | postgres | 1229 MB |
 public | interest         | table | postgres | 344 MB  |
 public | skill            | table | postgres | 438 MB  |
 public | skill_skill      | table | postgres | 21 GB   |
 public | title            | table | postgres | 1841 MB |
 public | title_interest   | table | postgres | 2799 MB |
 public | title_skill      | table | postgres | 27 GB   |
 public | title_title      | table | postgres | 11 GB   |
(13 rows)

正在查询的表格架构

cotitsk=# \d+ skill_skill
                      Table "public.skill_skill"
  Column   |  Type  | Modifiers | Storage | Stats target | Description
-----------+--------+-----------+---------+--------------+-------------
 skill1_id | bigint | not null  | plain   |              |
 skill2_id | bigint | not null  | plain   |              |
 count     | bigint | not null  | plain   |              |
Foreign-key constraints:
    "skill_skill_skill1_fk" FOREIGN KEY (skill1_id) REFERENCES skill(skill_id)
    "skill_skill_skill2_fk" FOREIGN KEY (skill2_id) REFERENCES skill(skill_id)

正在查询的表的近似行数

cotitsk=# SELECT reltuples::bigint AS estimate FROM pg_class where relname='skill_skill';
 estimate
-----------
 435104320
(1 row)

解释分析

cotitsk=# explain analyze select * from skill_skill where skill1_id = '2701941' order by count desc limit 1000;
                                                                QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------------
 Limit  (cost=8228458.83..8228461.33 rows=1000 width=24) (actual time=37163.205..37163.600 rows=1000 loops=1)
   ->  Sort  (cost=8228458.83..8229292.78 rows=333580 width=24) (actual time=37163.203..37163.394 rows=1000 loops=1)
         Sort Key: count DESC
         Sort Method: top-N heapsort  Memory: 127kB
         ->  Seq Scan on skill_skill  (cost=0.00..8210169.00 rows=333580 width=24) (actual time=14021.081..37128.913 rows=210902 loops=1)
               Filter: (skill1_id = '2701941'::bigint)
               Rows Removed by Filter: 434893298
 Planning time: 0.062 ms
 Execution time: 37163.747 ms
(9 rows)

postgresql.conf中

max_connections = 1000
shared_buffers = 16GB
work_mem = 100MB
maintenance_work_mem = 1GB
dynamic_shared_memory_type = posix
wal_buffers = 16MB
effective_cache_size = 48GB
logging_collector = on
log_filename = 'postgresql-%a.log'
log_truncate_on_rotation = on
log_rotation_age = 1d
log_rotation_size = 0
log_timezone = 'UTC'

Postgresql简单查询响应时间非常慢

0 个答案: