Re: Separating bgwriter and checkpointer

On Tue, Oct 4, 2011 at 2:51 AM, Dickson S. Guedes <listas(at)guedesoft(dot)net> wrote:
> 2011/10/3 Simon Riggs <simon(at)2ndquadrant(dot)com>:
>> On Sun, Oct 2, 2011 at 11:45 PM, Dickson S. Guedes <listas(at)guedesoft(dot)net> wrote:
>>> I'm trying your patch, it was applied cleanly to master and compiled
>>> ok. But since I started postgres I'm seeing a  99% of CPU usage:
>>
>> Oh, thanks. I see what happened. I was toying with the idea of going
>> straight to a WaitLatch implementation for the loop but decided to
>> leave it out for a later patch, and then skipped the sleep as well.
>>
>> New version attached.
>
> Working now but even passing all tests for make check, the
> regress_database's postmaster doesn't shutdown properly.
>
> $ make check
> ...
> ...
> ============== creating temporary installation        ==============
> ============== initializing database system           ==============
> ============== starting postmaster                    ==============
> running on port 57432 with PID 20094
> ============== creating database "regression"         ==============
> ...
> ============== shutting down postmaster               ==============
> pg_ctl: server does not shut down
> pg_regress: could not stop postmaster: exit code was 256
>
> $ uname -a
> Linux betelgeuse 2.6.38-11-generic-pae #50-Ubuntu SMP Mon Sep 12
> 22:21:04 UTC 2011 i686 i686 i386 GNU/Linux
>
> $ grep "$ ./configure" config.log
>  $ ./configure --enable-debug --enable-cassert
> --prefix=/srv/postgres/bgwriter_split

Yes, I see this also. At the same time, pg_ctl start and stop seem to
work fine in every mode, which is what I tested. Which seems a little
weird.

I seem to be having problems with HEAD as well.

Investigating further.

--
 Simon Riggs                   http://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services