Re: BUG #14378: when drop schema, standby will do many more than primary, and very slow, it's a bug?

Поиск
Список
Период
Сортировка
От 德哥
Тема Re: BUG #14378: when drop schema, standby will do many more than primary, and very slow, it's a bug?
Дата
Msg-id 67dc2404.2646.157da9f094f.Coremail.digoal@126.com
обсуждение исходный текст
Ответ на Re: BUG #14378: when drop schema, standby will do many more than primary, and very slow, it's a bug?  (Andres Freund <andres@anarazel.de>)
Список pgsql-bugs
Дерево обсуждения

HI,
   Add some info when i use perf top.  

1\. when drop schema on primary (search depend)
```
  findDependentObjects                           /home/digoal/pgsql9.6/bin/postgres
             3745.00 11.3% get_tabstat_entry                              /home/digoal/pgsql9.6/bin/postgres
```

2\. when drop schema start (drop files)
```
             5862.00 17.7% DropRelFileNodesAllBuffers                     /home/digoal/pgsql9.6/bin/postgres                                
             4058.00 12.2% smgrclose                                      /home/digoal/pgsql9.6/bin/postgres 
```

3\. when apply on standby
```
            12464.00 92.5% DropRelFileNodesAllBuffers                     /home/digoal/pgsql9.6/bin/postgres          
               56.00  0.6% DropRelFileNodeBuffers                         /home/digoal/pgsql9.6/bin/postgres  
```

best regards.


--
公益是一辈子的事,I'm Digoal,Just Do It.


At 2016-10-19 02:32:08, "Andres Freund" <andres@anarazel.de> wrote:
>On 2016-10-18 07:57:33 +0000, digoal@126.com wrote:
>> The following bug has been logged on the website:
>> 
>> Bug reference:      14378
>> Logged by:          Zhou Digoal
>> Email address:      digoal@126.com
>> PostgreSQL version: 9.6.0
>> Operating system:   CentOS 6.x x64
>> Description:        
>> 
>> 
>> 
>> when i use postgresql 9.6.0 for primary-standby, i drop a schema on primary,
>> but it's apply very slow on standby. 
>>   
>> the wal send very fast, but apply very slow, i think the syscall more than
>> it on primary? 
>> 
>> this is my test:
>
>There's some O(n^2) behaviour around lock release and acquire during
>replay IIRC.  See for example the thread around
>http://archives.postgresql.org/message-id/20140630231741.GA30918%40alap3.anarazel.de
>
>It's not entirely trivial to fix that though. E.g. the patch of Simon's
>later in that thread wouldn't help in your situation.
>
>Greetings,
>
>Andres Freund

В списке pgsql-bugs по дате отправления: