Re: [HACKERS] Why must -N be >= 16? - Mailing list pgsql-hackers

From Tom Lane
Subject Re: [HACKERS] Why must -N be >= 16?
Date
Msg-id 28712.947830966@sss.pgh.pa.us
Whole thread Raw
In response to Why must -N be >= 16?  (The Hermit Hacker <scrappy@hub.org>)
List pgsql-hackers
The Hermit Hacker <scrappy@hub.org> writes:
> Just trying to start up a 'mini-process' of -B 8 -N 4, and it tells me
> that -N  must be >= 16 ... why?  

You misread it --- -N can be as small as you like, but we don't allow
a really tiny -B.  To quote the code:
   if (NBuffers < 2 * MaxBackends || NBuffers < 16)   {       /* Do not accept -B so small that backends are likely to
starvefor        * lack of buffers.  The specific choices here are somewhat arbitrary.        */       fprintf(stderr,
"%s:-B must be at least twice -N and at least 16.\n",               progname);       exit(1);   }
 

I'm not even real sure that -B 16 is going to work well if you throw
complex queries at it --- we've not stressed the system with small
numbers of buffers.
        regards, tom lane


pgsql-hackers by date:

Previous
From: Mike Mascari
Date:
Subject: Re: Multiple Spindles ( Was: Re: [HACKERS][hackers]development suggestion needed )
Next
From: Tom Lane
Date:
Subject: Re: [HACKERS] SPI_fnumber can't see oid